Golgot13, thank you for explanation.
Can you list (here or private) model of BD players and # firmware which is able to play replicated (pressed) BD9 with menu, subtitles and at least 2 soundtracks?

First:
-AVCHD is not a BD9. AVCHD is like BD in HDMV on DVDR disc with AVC video codec and AC3 (1.0 to 5.1) or PCM audio track.
AVCHD is not in BDJ format (some player can accept BDJ but disc must start in HDMV). AVCHD is on DVDR so there is not AACS protection.
It's possible to do a AVCHD with PoP-up menu and many audio and subtitle track but there is many restrictions.
The big one is the muxed Data Rate, 30Mbps, => bitrate of video, audio, subtitle and popup menu (yes! popup menu
see the option properties of authoring tool for menu).

-BD9 is a BD (HDMV or BDJ) on DVD disc with some restriction. So there is AACS protection and Rom Mark (today the mastering
tool is compatible with "Rom Mark" protection on DVD, but there were not Rom Mark on first BD9 sample/test disc)
I recommand to read the BD Specification Rom 2 Part 1 and Part 2.
EclipseData mastering solution is fully compatible with BD9.



BD9 is on BD specification !!! But all BD players don't support it...
All Sony (include PS3) and Panasonic BD player can play BD9 and AVCHD. I have a feedback about Pioneer for BD9.
First BD player from Samsung or LG don't support it.


I think Mr Nazarian, DVDA, talked about AVCHD because there is not AACS.

Well me I say that this DVDForum test is completely ridiculous like psnr ratio for quality equivalence from microsoft.
As they say, there is always a first time for everything. Here we have someone who dismisses subjective and objective tests! Don't know that I have seen anyone be in both camps at once!

Or maybe you are against any test in which your favorite codec did not do well ;).
show me the "real" source of your choice ... and make the encoding. Show that VC1 is never close that to MPEG2 at 3:1 ratio (for objective or sujective test) will be really simple for me. I doubt that double blind with golden eyes will be necessary to show cleary that ...
I don't work for Microsoft but even if I did, the only side of me you see would be my back as I ignored the random challenges from you two :). As I have said time and time again, VC-1 has proven itself and needs no vote of confidence from the likes of people making challenges of this sort.
And I have perhaps an explanation for DVDForum test result. MPEG2 encoding use certainely HDDVD spec here. HDDVD spec are 29.4 to max bitrate (perhaps less for this test). Use 24 Mbps with 29.4 Mbps imply in practice really constrained VBR mode. In practice really close to CBR mode. MPEG2 is clearly unadapted for HDDVD (30 GB and max at 29.4 Mbps). It's not the case with BD with max rate at 40 Mbps and 50 GB.
Looks like you are still in the dark as to nature of DVD forum tests and are just throwing random guesses out there to see what sticks. So let me help you out a bit. There were two versions of MPEG-2 in DVD Forum tests. One at 24 mbit/sec and another at just 7.7 mbit/sec. At 7.7 mbit/sec MPEG-2 beat AVC on a number of tests. Do you have an explanation for that? Yes, I will give you time to go and consult with your codec friend :).

What is this filteing pre-process? For many people it would sort of limit the value of a test if it was removing any high frequency detail.

- Tom

It's dark noise filtering:
- Use really agressive denoising for luma Y < 32
- Use soft denoising to reduce temporal complexity for luma Y > 32

It's really well know HVS pre-process. It's really hard for eyes to see grain/noise in dark area. Grain/noise imply high entropy. Always difficult for codec to code that. You have many advantage to make that:

1) Denosing in dark area save bit for use in another high luma area.
2) You can use AQ to improve quant in dark area.


source=crop(source,0,140,-0,-140)

sourceA=FFT3dfilter(source,sigma=0.1,plane=4)
sourceB=FFT3dfilter(source,sigma=2.5,plane=4)

th_low = 0
th_high = 32

dmask = sourceB.levels(th_low,1.0,th_high,255,0,false)

result=sourceA.mt_merge(sourceB,dmask,U=3,V=3,luma=true)

result=addborders(result,0,140,0,140)
return result

As they say, there is always a first time for everything. Here we have someone who dismisses subjective and objective tests! Don't know that I have seen anyone be in both camps at once!

Or maybe you are against any test in which your favorite codec did not do well ;).

Well H264 is not my favorite codec. I have just the best result with H264. I return the question ... lol



Looks like you are still in the dark as to nature of DVD forum tests and are just throwing random guesses out there to see what sticks. So let me help you out a bit. There were two versions of MPEG-2 in DVD Forum tests. One at 24 mbit/sec and another at just 7.7 mbit/sec. At 7.7 mbit/sec MPEG-2 beat AVC on a number of tests. Do you have an explanation for that? Yes, I will give you time to go and consult with your codec friend :).

... and never beat VC1 at 7.7 Mbps I suppose ... lol

Yes I have explanation. And it's a confirmation. Really really ridiculous test. DVDForum use completely crappy implementation like microsoft psnr internal test ... lol


Stop trolling and make encoding ... with the source of your choice. I'am really curious to see that.

Yes I have explanation. And it's a confirmation. Really really ridiculous test. DVDForum use completely crappy implementation like microsoft psnr internal test ... lol
Well, that is a shame then. Because the explanation could have helped your cause! And the answer had nothing to do with the test but the design of the codec.

Well, that is a shame then. Because the explanation could have helped your cause! And the answer had nothing to do with the test but the design of the codec.

I'm interested to know the answer as to why MPEG 2 was better than AVS at 7.7mbps in some tests?

Well, that is a shame then. Because the explanation could have helped your cause! And the answer had nothing to do with the test but the design of the codec.

Well, I for one am curious. Why did MPEG-2 beat AVC on some of the 7.7 mbps tests?

- Tom

Regarding filtering: Why do you filter the source material *at all* for this codec challenge?

I thought the dream of the users reading/posting on this board is and always has been to stay as close to the (uncompressed) source as possible. Ideally, an analog film source would be captured in 2k or 4k, and then sent to the encoder without interim modification of any kind, to make sure everything what's on the film source gets preserved and displayed in our home theaters, exactly as we've seen it on the cinema screen.

So Zodiaque, as Trbarry asked: why not take a real uncompressed unfiltered source, let the encoders do their work, and then compare the results with the original?

So Zodiaque, as Trbarry asked: why not take a real uncompressed unfiltered source, let the encoders do their work, and then compare the results with the original?

Because in the real encodind world all the uncompressed source are always pre-processed to help the codec during the encoding. My filtering is in practice transparent for the eyes. You can't see difference between original source and filtered source. Anyway codec "see" real difference. Here I use the same typicall pre-process that VC1 use for HDDVD or BD encoding.

Well, I for one am curious. Why did MPEG-2 beat AVC on some of the 7.7 mbps tests?

- Tom

Yes ... me too. Really.

But I doubt seriousely that amirm show this source.

blablablablabla. Amirm is the best to speak. Unfortunaly never show clear example or demonstration. I wait always the 1080p24 uncompressed source from benwaggoner for example.

I can make many encoding with real source that show cleary that:
- H264/VC1 never beat or even on par at 1:3 ratio (subjectively or objectively)
- VC1 never beat H264 for objective test.
- H264 subjectively better than VC1 for a large majority of source.

I can make many encoding with real source that show cleary that:
- H264/VC1 never beat or even on par at 1:3 ratio (subjectively or objectively)

Here comes a question. Is 25mbs a must for mpeg2 or can a skilled compressionist achive the same quality with lower bitrate like 18 to 20 mbs?

Here comes a question. Is 25mbs a must for mpeg2 or can a skilled compressionist achive the same quality with lower bitrate like 18 to 20 mbs?

Compressionist are generaly bad encoder expert. Generaly they doesn't understand very well the available setting for the encoder.

First at all compressionist want:
- Compliance
- Speed

Quality is generaly not the first priority.

My point I was refering to.

Many says that its impossible for VC1 clip at 8mbs to look near as good as a mpeg2 at 25mbs.

But that assumption is based that the Mpeg2 clip was optimized at its 25mbs running speed.

If you could reach the same quality at 18mbs mpeg2, then would the VC1 clip be near 2X the efficency.

So that a 8mbs VC1 clip did so well compared to a 3X Mpeg2 clip, doesnt equal that VC1 is 3 times more efficient. Just that it behaved very well despite having 3X less the bitrate.

The Lion is incorrect .... as usual. Most of Nature's Journey was shot with the following Professional Camera.

Sony 750 HDCAM 1920x1080 Camera with a $20,000 Cannon Lens.

This is NOT a "Prosumer" camera.

Only a few shots used the "Prosumer" cameras. They were only used when we were dealing with locations in physically challenging places where we could not easily haul-in the Sony 750.

Also, 99% of people who have reviewed this title say quite the opposite of what he says. Have you yourself seen Nature's Journey? Does it in any way look soft to you?

If it looks soft to The Lion, maybe it is because of the VC-1 CODEC we used ...:)

It is funny how the MS Minions will attack anything, everything, and everyone that will not fall under the spell of MS ... even going as far as eating thier young by attacking a title that uses VC-1.

This is why I no longer visit this Forum very much.

Ah! So Nature's Journey truly is shot with interlaced video cameras and not in progressive or with film. Now I remember the comments at the time that led me to skip this purchase.

I found this quote (http://www.avsforum.com/avs-vb/showthread.php?p=11959753#post11959753)from TheLion:


Wasn't there a lot of work done later on VC-1 to make it work well with rock concert video? Perhaps it was too late for Nature's Journey .

Since Nature's Journey uses interlaced video as an original source, the codec discussion could focus on what video codecs set to output 1920x1080 results, do well with that type of source. What encoder settings should be used? What different settings would be used for BD target versus a downloadable format targeting a media player that skips going to optical disc?

The Lion is incorrect .... as usual. Most of Nature's Journey was shot with the following Professional Camera.

Sony 750 HDCAM 1920x1080 Camera with a $20,000 Cannon Lens.

This is NOT a "Prosumer" camera.
If it was captured HDCAM, then it was anamorphic 1440x1080 with 3:1:1 color. So at best you had only 75% of the horizontal luma samples and 50% of the horizontal chroma samples of a full 1920x1080 4:2:0 or 4:2:2 source.

A more likely source for loss of detail...

Well, I for one am curious. Why did MPEG-2 beat AVC on some of the 7.7 mbps tests?

In what year it was? In same when microsoft said VC1 in 3 times better than mpeg2? :)

I'm interested to know the answer as to why MPEG 2 was better than AVS at 7.7mbps in some tests?
The answer is in archives of this forum actually. Let the question soak for a while to see if Zodiaque has anything of value to say about it other than more poorly understood assumptions.

If it looks soft to The Lion, maybe it is because of the VC-1 CODEC we used ...:)

It is funny how the MS Minions will attack anything, everything, and everyone that will not fall under the spell of MS ... even going as far as eating thier young by attacking a title that uses VC-1.
TheLion is the last person on this forum to be a "MS Minion." Maybe you want to go and read some of the heated discussions between him and me. So if he says what he says, it isn't motivated by any love for me personally, or the company.

As to VC-1 being at fault, did you not compare the output to the source before you were done? If it didn't perform, why not go and use MPEG-2 which you praised earlier?

Besides, in the email you quoted from the third-party who did the encoding, there is a lot of talk about poor source material. How is it that you say that the codec was more at fault when your own post indicates major issues with the *source*?
This was very difficult source material to encode and it took a lot of re-encoding and some source redeliveries to make the picture look good and to pass the Verifiers. Some of the difficulties were caused by the codec itself trying to deal with the source material. For instance, the source had a lot of recorded in Macro blocking that proved difficult for the codec to handle. In addition, because of the challenges with the source material, we had to carefully manage the re-encode settings in order to avoid buffer allocation errors that were being flagged by the verifiers.

If you feed the codec a source with blocking artifacts, then the codec will attempt to preserve those high frequency edges and make the rest not look as good. Or it may make those edges look worse. I can't imagine HDCam material having so many issues. Did someone pre-compress the material before giving it to the third-party? If the source was digital to start, what did you do different in the re-submission of them? Filter the artifacts? Really puzzling to read comments from your compressionists regarding such source artifacts on an eye candy source.

This is why I no longer visit this Forum very much.
Well, I hope you hang around enough for us to understand issues above which you should be able to answer as the comments regarding poor condition of the source came from your side of the fence.

Finally, with the format war over, I hope you can put aside the tone of your commentary as shown in this post. It doesn’t add any information to the forum. And only serves as a trigger for folks to find holes in your arguments as shown here :).

Compressionist are generaly bad encoder expert. Generaly they doesn't understand very well the available setting for the encoder.

First at all compressionist want:
- Compliance
- Speed

Quality is generaly not the first priority.

Since the compressionists are resonsible for the end product and you're saying they "suck" why even argue about codecs? Clearly, the bottleneck here isn't the codec but the compressionst, according to you.

One day I'm hoping cloning is legalized and we can finally replace these bastard compressionsts with little Zodiaque's who'll give us the bestest quality ever and we can all be happy.

Maybe this topic should be retitled to the effect of "Compressionist Wars", rather than "Codec Wars"? :D

Maybe this topic should be retitled to the effect of "Compressionist Wars", rather than "Codec Wars"? :D

You need both to make an encode.

I wish I was superrich, becaused then I could finance the biggest codecshootout the earth has ever seen,
with all the members of AVS forums as test subjects.

Could be fun. :)

Since Nature's Journey uses interlaced video as an original source, the codec discussion could focus on what video codecs set to output 1920x1080 results, do well with that type of source. What encoder settings should be used? What different settings would be used for BD target versus a downloadable format targeting a media player that skips going to optical disc?
I happen to have some eye candy 1080i source if we'd like to look at some samples of that in BD specs.

Here's a QHD sample of it:

http://www.on10.net/blogs/benwagg/Sample-Encoder-Test-Clips/

Early "draft" version of the 24p encode that Golgot13 keeps asking about, but never seems to have any commentary on is there as well.

Compressionist are generaly bad encoder expert. Generaly they doesn't understand very well the available setting for the encoder.

First at all compressionist want:
- Compliance
- Speed

Quality is generaly not the first priority.
Perhaps your industry has much lower standards, but that's manifestly untrue of the Hollywood compressionists in charge of making studio Blu-ray titles.

It's really well know HVS pre-process. It's really hard for eyes to see grain/noise in dark area. Grain/noise imply high entropy. Always difficult for codec to code that. You have many advantage to make that:
Eh?

Perhaps in the CRT era with perceptually uniform gamma, or in that mythical land where every user calibrates their display perfectly (and hence watches in a dark room).

But in the Plasma/LCD era, blocking and noise in black is a major problem, and arguably has been the single biggest place we've spent time in our VC-1 development. Studio compressionists run "consumer calibrated" displays with elevated blacks just to make sure that they AREN'T getting a bunch of problems in there.

If you are running displays calibrated so you don't see black detail, and filtering that out aggressively, that could account for a lot of the difference between codecs you're seeing. But you're also going to make content that's going to be problematic on typical displays with typical calibration.

We spent a lot of bits in PEP's VC-1 to make sure we preserve black detail.

Eh?

Perhaps in the CRT era with perceptually uniform gamma, or in that mythical land where every user calibrates their display perfectly (and hence watches in a dark room).

But in the Plasma/LCD era, blocking and noise in black is a major problem, and arguably has been the single biggest place we've spent time in our VC-1 development. Studio compressionists run "consumer calibrated" displays with elevated blacks just to make sure that they AREN'T getting a bunch of problems in there.

If you are running displays calibrated so you don't see black detail, and filtering that out aggressively, that could account for a lot of the difference between codecs you're seeing. But you're also going to make content that's going to be problematic on typical displays with typical calibration.

We spent a lot of bits in PEP's VC-1 to make sure we preserve black detail.

From PEP and PSE documentation ...

Recommended settings for re-encoding

You can make several settings changes for the third encoding pass to improve
problematic segments. The following sections discuss recommendations for
changes you can make.

Control dark levels by using the dark region denoise filter
Selecting Enable dark region denoise improves the quality in dark areas of the
video image, particularly dark regions that exhibit a lot of film grain. You have
several options for how to apply the dark region denoise filter. You can choose
to apply a low-pass filter to reduce noise or you can use a flattening filter for
luminance, for chrominance, or for both. You can also set a value for how much
processing, or strength, the filter applies to the dark regions.


To fix blocking in black regions:

• Use the dark region denoise filter by selecting Enable dark region denoise
on the Re-Encode tab. You have several choices for how to apply the filter
and you can adjust its strength.

• Enable differential quantization by selecting Enable DQuant on the Re-
Encode tab. First, try changing the Aggressive Bit Usage settings for P and B
frames. Next, adjust the quantization settings for dark regions to force the
encoder to use additional bits in these areas by lowering the relative or
maximum QP, or by increasing the relative and minimum Quality settings.


I make exactly that for the Darknight source with H264, MPEG2 and VC1:
- Denoising in dark area ( Y < 32)
- AQ with better quality for low complexity area or/and dark area

Imply really better quality and details conservation in dark area

The answer is in archives of this forum actually. Let the question soak for a while to see if Zodiaque has anything of value to say about it other than more poorly understood assumptions.

Well I wait always your proposition for the source ... lol. Choose like want. I can't say more. But I think that you know very well the result, isn't it? For example with this Darknight source there are no doubt.

I make exactly that for the Darknight source with H264, MPEG2 and VC1:
- Denoising in dark area ( Y < 32)
- AQ with better quality for low complexity area or/and dark area

Imply really better quality and details conservation in dark area
Sure, if you have actual non-signal noise in the black areas, that's a source issue worth fixing. But you need to be able to preseve real black detail that isn't noise.

What MPEG-2 encoder are you using?

Also, did the source have grain or noise in the low blacks?

Well I wait always your proposition for the source ... lol.
Then you will wait forever since you don’t seem to read anything I tell you about your challenge.
But I think that you know very well the result, isn't it?
Oh I know the results too. If AVC loses, then you will call the source/testing ridiculous :p.

Really, this discussion has become so information-free. When I ask you to engage on a technical matter of evaluating test methodology for DVD Forum or other technical detail, all I read is “LOL.” What is that supposed to mean? This is not a chat on AOL for heaven’s sake. Bring some data and know-how to the thread so for people who don’t care about the specific clip/challenge of yours, they learn something new. Without it, it seems like a childish brawl and nothing more.

Sure, if you have actual non-signal noise in the black areas, that's a source issue worth fixing. But you need to be able to preseve real black detail that isn't noise.

FFT3dfilter(source,sigma=2.5,plane=4) is agressive for noise/grain but not for details. Details conservation is good with this sigma.


What MPEG-2 encoder are you using?

Well I make test encoding with libavcodec, Mainconcept, HCEnc, Procoder and many other. For this source I obtain best result with the old ... TMPGEnc.

Then you will wait forever since you don’t seem to read anything I tell you about your challenge.

Oh I know the results too. If AVC loses, then you will call the source/testing ridiculous :p.


Well the encoding will be available ... users can see themselves the result. It's really strange: you can choose the source, the methodology, make the VC1 encoding ... and you refuse simply to indicate a source. you seem to not be very confident for the result!



Really, this discussion has become so information-free. When I ask you to engage on a technical matter of evaluating test methodology for DVD Forum or other technical detail, all I read is “LOL.” What is that supposed to mean? This is not a chat on AOL for heaven’s sake. Bring some data and know-how to the thread so for people who don’t care about the specific clip/challenge of yours, they learn something new. Without it, it seems like a childish brawl and nothing more.

lol ... ;-)

Well the encoding will be available ... users can see themselves the result. It's really strange: you can choose the source, the methodology, make the VC1 encoding ... and you refuse simply to indicate a source. you seem to not be very confident for the result!
I just re-posted a link to some proposed content upthread. Do YOU have any thoughts on them?

I just re-posted a link to some proposed content upthread. Do YOU have any thoughts on them?

where is the link?

where is the link?
http://www.avsforum.com/avs-vb/showpost.php?p=14260866&postcount=773

http://www.avsforum.com/avs-vb/showpost.php?p=14260866&postcount=773

Well why not. I have 10 GB ftp for you if you want. With lossless codec like logarith it will be possible to store ~12 min on 10 GB.

Well why not. I have 10 GB ftp for you if you want. With lossless codec like logarith it will be possible to store ~12 min on 10 GB.
I'm asking you to LOOK at the proposed clips, and suggest what if anything they'd need to be useful for codec testing.

I'm asking you to LOOK at the proposed clips, and suggest what if anything they'd need to be useful for codec testing.

Seem good like that. Difficult to evaluate the compressibility of this source like that. What bitrate(s) do you want use for this encoding?

I'm asking you to LOOK at the proposed clips, and suggest what if anything they'd need to be useful for codec testing.

I have no interest in the 1080i clips for a shootout purpose. And I personally don't have any comments or suggestions about the 24p source clip as long as any sort of consensus can be reached in some finite period of time.

But getting untouched 24p film source for a shootout does seem to be a bit difficult. ;)

- Tom

It's really strange: you can choose the source, the methodology, make the VC1 encoding ... and you refuse simply to indicate a source. you seem to not be very confident for the result!
What is really strange is that I have explained this half a dozen times. Even others have tried to explain it to you. Yet, here you go again sounding like I am talking to a blank wall. In the words of my Andy Dufresne in my favorite movie Shawshank Redemption "How can you be so obtuse? Is it deliberate?" ;) :)


lol ... ;-)
Gosh, another information free post. And I am making it worse by responding in kind....

What is really strange is that I have explained this half a dozen times. Even others have tried to explain it to you. Yet, here you go again sounding like I am talking to a blank wall. In the words of my Andy Dufresne in my favorite movie Shawshank Redemption "How can you be so obtuse? Is it deliberate?" ;) :)



Gosh, another information free post. And I am making it worse by responding in kind....

The more you encourage him.....

I'm asking you to LOOK at the proposed clips, and suggest what if anything they'd need to be useful for codec testing.

1080p source will be good ...

@ Zodiaque

I downloaded only the 12Mbps AVC clip so far.
I see that there is too much banding (on my Eizo 2231 slightly calibrated screen)
a. @ 10sec (fading logos)
b. @ 23-28sec (deep blue night Sky).

I wonder .... is it (the banding) there in the prototype source ?? If yes, isn't it possible to take care of it (dithering). Or is it a side effect of your pre proccess. Or of the avc encoding maybe ??

I have no interest in the 1080i clips for a shootout purpose. And I personally don't have any comments or suggestions about the 24p source clip as long as any sort of consensus can be reached in some finite period of time.

But getting untouched 24p film source for a shootout does seem to be a bit difficult. ;)

It is. WHich is why Microsoft did that shoot that resulted in the 1080i and 1080p clips. However, it's fallen on me as the only guy interested in finishing up the 1080p. And since I'm really not an editor, I need all the help I can get :).

I downloaded only the 12Mbps AVC clip so far.
I see that there is too much banding (on my Eizo 2231 slightly calibrated screen)
a. @ 10sec (fading logos)
b. @ 23-28sec (deep blue night Sky).

I wonder .... is it (the banding) there in the prototype source ?? If yes, isn't it possible to take care of it (dithering). Or is it a side effect of your pre proccess. Or of the avc encoding maybe ??
Banding is in the source.

Luma levels also don't match between letterboxing and the actual image - elevated in the image area.

It is. WHich is why Microsoft did that shoot that resulted in the 1080i and 1080p clips. However, it's fallen on me as the only guy interested in finishing up the 1080p. And since I'm really not an editor, I need all the help I can get :).

Hey, I'm even further from being an editor since I'm a bits & bytes guy with the genetically built-in talent of making things that are technically functional but aesthetically sort of bizarre and nerdy.

But we can offer moral support. :cool:

And I think Microsoft could gain points here by participating, regardless of the final results.

- Tom

I have 1080P content you guys can use...who needs what for what shootout?

I own all worldwide rights to my content so I can grant you use for a test.

I have no interest in the 1080i clips for a shootout purpose. And I personally don't have any comments or suggestions about the 24p source clip as long as any sort of consensus can be reached in some finite period of time.

But getting untouched 24p film source for a shootout does seem to be a bit difficult. ;)

- Tom

Notice the smiley face....the VC-1 crack was joke.

We did compare...it looked "good enough" ...:) This too is a joke...as it looked very good when we ere done with the encode.

There were not "Major Issues" ... read it again. They said it was difficult material to encode but we corrected and redelivered any scenes that gave the encoders grief. In most cases, the problem was not the source, but the limitations of 8 Bit Color which causes banding in some scense .. as it would with any source .. and the inability of the encoders to digest some of the material. We had to go back and rework some content before the encoders would stop puking on it.

Also, there were some scenes where the creator ... for whatever artistic reasons he may have had ... that appeared to have macro blocking. This could have been due to the layering but I will ask him to be sure.

If you want to see the difference between a lower bit rate encode on HD-DVD ... verses what we did .... look at the 2008 DTS Demo Disc. They used Nature's Journey (2 Clips) and did the encodes at a lower rate to fit DTS Lossless Audio. I am pretty sure they used VC-1....if I am not mistaken.

TheLion is the last person on this forum to be a "MS Minion." Maybe you want to go and read some of the heated discussions between him and me. So if he says what he says, it isn't motivated by any love for me personally, or the company.

As to VC-1 being at fault, did you not compare the output to the source before you were done? If it didn't perform, why not go and use MPEG-2 which you praised earlier?

Besides, in the email you quoted from the third-party who did the encoding, there is a lot of talk about poor source material. How is it that you say that the codec was more at fault when your own post indicates major issues with the *source*?


If you feed the codec a source with blocking artifacts, then the codec will attempt to preserve those high frequency edges and make the rest not look as good. Or it may make those edges look worse. I can't imagine HDCam material having so many issues. Did someone pre-compress the material before giving it to the third-party? If the source was digital to start, what did you do different in the re-submission of them? Filter the artifacts? Really puzzling to read comments from your compressionists regarding such source artifacts on an eye candy source.


Well, I hope you hang around enough for us to understand issues above which you should be able to answer as the comments regarding poor condition of the source came from your side of the fence.

Finally, with the format war over, I hope you can put aside the tone of your commentary as shown in this post. It doesn’t add any information to the forum. And only serves as a trigger for folks to find holes in your arguments as shown here :).

*ahem*

Who is saying there is something wrong with the two VC-1 encodes of Nature's Journey?

Other than the material is just soft looking. Its artisitic intent BTW ;)

Right?

Well Richard Casey is offering 1080p material for the shootout... will we finally, after 799 posts get to see something? :)

In addition to the straight shootout between VC-1 and AVC, I'd also like to see how each perform at 12, 15, 20, and 30 mbps (may need to use shorted clips though)

It appears that some of the "AVS Folks" are still going back and taking bits out of context. For example, the reference by "the Lion" who said Nature's was all done with crappy consumer cameras ... this is not at all the case. There were only certain scenes that were shot with lower end "Prosumer" Cameras and it was typically due to the lack of easy access to the location. A great deal of the program was shot with a Sony 750 with a $20,000 lens.

As I will explain, this still does not make the source material perfect by any means. This actually supports my argument from the beginning that Digital is NOT perfect....far from it. I have yet to see any digital technology that can capture ultra fidelity picture and sound when it come to the recording of our analog world.

However, in the end, I truly believe that taking posted information out of context and spreading misinformation may actually be ... in part .... intentional in these Forums. Sadly it seems, this is the case more often than not. It also appears that the amount of misinformation may actually be tied to who and what you support or believe in.

Bottom line ... Digital is NOT perfect and the problems at the source are due to technical limitations of current HD Pro equipment such as 8 bit color, digital processing, and compression ... which is never kind to any A/V Source no matter how much certain people may try to convince you that it does not impact quality and that it is "good enough" ... I can assure you it is not. Lossy Compression, while a reality for practical purposes, is one of the worst things you can do to any A/V source.

Anyhow, to answer the question about the source content on Nature's Journey, I spoke to the original artist. Here are some comments which may clear things up ... or more likely cause further debate .... regarding the quality of the source material.

Again, only a portion of the program was shot with the FX-1 HDV. Funny thing ... even experts often say the FX-1 HDV Footage looks sharper that the Sony 750 footage. Go figure.

------

Macroblocking occurs from some of the footage shot with the FX-1 HDV camera in soft color areas like shades of blue around a cloud. The compression in the camera is limited in terms of color and therefore a group of pixels (usually 16x16) is the same color instead of each pixel being unique. This can be as small as 4x4 pixels.

(R&B Note - Only a portion of the footage was shot with the FX-1. However, even the Sony camera is not perfect. No digital camera is)

Deep shadows are also a culprit for this artifact. I often pushed the shadows up to get more detail and less apparent contrast but revealed some extra noise. To fight this, the scenes were often frame averaged (if there was no camera movement). Frame averaging evens out things ... usually over 5 or so frames ... and removes most macroblocking. Any compression system (including the sony 750 or 950) would occasionally do macroblocking as it's also an error in encoding and may be like black dropout, but it's more the limited color space of the hdv shots.

(R&B Note - The camera's compression and color space were the limiting factors and culprits here ... even with the more expensive Sony which was used for most of the shooting)

Again, let me say that the process I did often addressed this issue by adding multiple layers of composited contrast and additional RAW digital sources, but it's far far far from perfect.

(R&B Note - The artist and Technicolor reworked the content to minimize any artifacts or issues with the source material to minimize encoding artifacts and make the best possible presentation. Far more work went in to Nature's Journey to assure it was the best possible HD / BD presentation possible than anyone here on this forum realizes)

One has to remember that I have been slammed by people who saw dirt on the lens and the limited compression in the opening shot in Nature's when in actuality that is entirely computer generated and at full 4:4:4 uncompressed 10 bit color.

(R&B Note - This is a typical AVS Comment that is way off base. There was no lens used at all used in the opening scene. The source was the highest quality possible by today's standards ... yet people claimed it was no good becasue there was dirt on the lens ... a lens that never existed to begin with)

So here is the thing. In terms of encoding, it seems that the AVS universe thinks it's only valid for perfect imagery. If you use imperfect images then encoding is also at fault. Of course this makes no sense.

(R&B Note - I agree ... my whole point is that Digital is NOT perfect. Far from it. None of it is really capable of presenting a flawless representation of our analog world. Limitation range from limited color depth and processing artifacts to lossy compression and limited latitude)

Actually, noisy over sharpened images with lots of movement should be some of the hardest to encode. Technocolor deserves accolades for making some of the scenes look even better than the original, and they certainly maintained the nature of the images with great precision.

(R&B Note - Nobody here at AVS has any insight in to the entire production process for AVS except the Artist, R&B Films, and Technicolor. Yet they try to position themselves as experts on what is wrong with the production. Funnier still is that fact that many AVS Folks that believe they are experts and come across as "know it alls" regarding the production of not just Nature's Journey ... but High Definition content in general. That fact is that many, if not most, have never even produced HD Content from start to finish nor are they familiar with ... nor will they admit .... the serious limitations and challenges that once faces when working with today's HD technology. It is far from perfect...and that is fact as well as the bottom line)

The fact that AVS Folks are still attacking the source footage is just tiresome. It seems that folks there still think interlaced is half the resolution of progressive. They don't get that progressive is 29.97 frames whereas interlaced is 59.94 fields per second. Interlaced has it's own issues but resolution is not one of them.

(R&B Note - A lot of people here on the AVS Forum would like you to believe otherwise and they would also like you to tell you what is good and what is not good and what is "good enough" ... sorry ... but I would like to be the one who decides what is "good enough" for my taste and acuity)

Bottom line - Nature's Journey was produced by some of the most knowledgeable people in the industry ... hand picked by me ... to assure the best presentation possible of the Artists work. Whether or not one appreciates the creator's artistry is one thing ... however, anyone familiar with the entire production process would certainly appreciate the brilliance, skills, talent, and effort of everyone involved in the creation of this program.

I am closing this loop on Nature's Journey as all that could be said has been said at this point. I am going back to developing a new project that will either resolve or fuel many of the conflicts I see people going back and forth on here at AVS. It will be entertaining .... to say the least .... to watch all the so called "experts" crawl out of the woodwork to attack this production as well .... regardless of whether or not they have any facts about the production.

Besides, in the email you quoted from the third-party who did the encoding, there is a lot of talk about poor source material. How is it that you say that the codec was more at fault when your own post indicates major issues with the *source*?

If you feed the codec a source with blocking artifacts, then the codec will attempt to preserve those high frequency edges and make the rest not look as good. Or it may make those edges look worse. I can't imagine HDCam material having so many issues. Did someone pre-compress the material before giving it to the third-party? If the source was digital to start, what did you do different in the re-submission of them? Filter the artifacts? Really puzzling to read comments from your compressionists regarding such source artifacts on an eye candy source.


Well, I hope you hang around enough for us to understand issues above which you should be able to answer as the comments regarding poor condition of the source came from your side of the fence.

Finally, with the format war over, I hope you can put aside the tone of your commentary as shown in this post. It doesn’t add any information to the forum. And only serves as a trigger for folks to find holes in your arguments as shown here :).

I am more than willing to provide a master of CHRONOS on D5. It is not impeccable source... but it is real world ... which is what an encoder test should be based on.

I am also willing to provide segments of Nature's Journey to see if anyone could do better that we could with the encodes. My money would bet on the "no pass" line.

All I would insist on is that we supervise the encodes and do them at a a place we all agree is neutral. Each part could work with the compression folks to assure the fairness of the test.

Well Richard Casey is offering 1080p material for the shootout... will we finally, after 799 posts get to see something? :)

In addition to the straight shootout between VC-1 and AVC, I'd also like to see how each perform at 12, 15, 20, and 30 mbps (may need to use shorted clips though)

I am more than willing to provide a master of CHRONOS on D5. It is not impeccable source... but it is real world ... which is what an encoder test should be based on.

Yes a D5 of Chronos would be more suitable since it would be equal for both codecs.

Interlaced has it's own issues but resolution is not one of them.
While I appreciate your sentiments to clear up much of the confusion around 1080i vs 1080p, this statement is not going to do that :). Interlace most definitely impacts resolution in vertical domain.

If you have a still or progressive image, then interlace is indeed the same as progressive. This is why 1080i (movie) output of BD/HD DVD is the same as 1080p. The source is not changing between fields so whether you send an entire frame or half at a time, doesn't make a difference.

However, if the image is moving, then its vertical resolution starts to drop and at full motion, it goes way down from 1080 pixels to 540. Horizontal resolution remains the same. Reason is simple: each field only has 540 pixels of vertical resolution. So if the fields are different from each other due to motion, then each represents a different image and when put back together in the display, does not recreate the 1080 lines as we would be the case with still image or progressive sources.

Put another way, interlace is a compression method whereby the vertical resolution is adaptively reduced depending on level of motion in the image. For still images, we get all the benefits of this compression without penalty (compression is lossless). For moving images, then we still get the data reduction, but in return, get "compression artifacts" in the form of jaggies (compression is lossy).

So in the case of Nature's Journey shot on interlace camera to start, the segments with subject/camera motion are NOT at the same resolution as 1080p. They vertical resolution is decimated for good at the time of capture. You can try to use different methods to get rid of the distortion introduced but you are going to have a hard time matching true 1080p camera. For example, if the motion cannot be tracked properly, even a motion adaptive de-interlacing system is going to resort to filtering and as such, have less than 1080 lines of vertical resolution.

While I appreciate your sentiments to clear up much of the confusion around 1080i vs 1080p, this statement is not going to do that :). Interlace most definitely impacts resolution in vertical domain.

If you have a still or progressive image, then interlace is indeed the same as progressive. This is why 1080i (movie) output of BD/HD DVD is the same as 1080p. The source is not changing between fields so whether you send an entire frame or half at a time, doesn't make a difference.

However, if the image is moving, then its vertical resolution starts to drop and at full motion, it goes way down from 1080 pixels to 540. Horizontal resolution remains the same. Reason is simple: each field only has 540 pixels of vertical resolution. So if the fields are different from each other due to motion, then each represents a different image and when put back together in the display, does not recreate the 1080 lines as we would be the case with still image or progressive sources.

Put another way, interlace is a compression method whereby the vertical resolution is adaptively reduced depending on level of motion in the image. For still images, we get all the benefits of this compression without penalty (compression is lossless). For moving images, then we still get the data reduction, but in return, get "compression artifacts" in the form of jaggies (compression is lossy).

So in the case of Nature's Journey shot on interlace camera to start, the segments with subject/camera motion are NOT at the same resolution as 1080p. They vertical resolution is decimated for good at the time of capture. You can try to use different methods to get rid of the distortion introduced but you are going to have a hard time matching true 1080p camera. For example, if the motion cannot be tracked properly, even a motion adaptive de-interlacing system is going to resort to filtering and as such, have less than 1080 lines of vertical resolution.

And most decent deinterlacing algorithms I know about will probably also end up filtering somewhat horizontally too, to get rid of the stair step jaggies caused by vertical aliasing of trying to put more than 540 vertical pixels of information in each field. Meanwhile these days, in a world of fixed pixel displays all 1080i is deinterlaced.

- Tom

I am closing this loop on Nature's Journey as all that could be said has been said at this point. I am going back to developing a new project that will either resolve or fuel many of the conflicts I see people going back and forth on here at AVS. It will be entertaining .... to say the least .... to watch all the so called "experts" crawl out of the woodwork to attack this production as well .... regardless of whether or not they have any facts about the production.

In 1080p with no overprocessed, filtered look like the last time I hope :)

Make it Au Natural ;)

It seems that folks there still think interlaced is half the resolution of progressive. They don't get that progressive is 29.97 frames whereas interlaced is 59.94 fields per second
While I appreciate your sentiments to clear up much of the confusion around 1080i vs 1080p, this statement is not going to do that :). Interlace most definitely impacts resolution in vertical domain.

So in the case of Nature's Journey shot on interlace camera to start, the segments with subject/camera motion are NOT at the same resolution as 1080p

Amir, while I agree there are issues with interlacing, and while I agree that per refresh, interlaced has lower resolution than progressive, you're directly comparing 1080i and 1080p without taking into account the refresh (frame or field) rate.

Most films are 1080p24 but Nature's Journey is 1080i60. I agree 1080i60 will only have 540 fields per refresh but once both have been put on screen it should give more full frames per sec than 1080p24 (ie. around 1080p30 - kind of) - assuming the player/display doesn't filter them and just puts them on-screen above/below the previous field (if that's how it works and it doesn't double each field).

If you're filming a tiger running at 50mph the 1080i60 rate should capture the motion a lot more accurately than 1080p24.

Also, why aren't more things filmed at 1080p30 (and stored at 1080i60)? That's progressive (stored as interlaced but displayed as 30p progressive), would have more accurate motion portrayal than 1080p24 (though not as good as 1080i60) but not have the interlacing artefacts, and it should still retain the film-look. And until we have a consumer playback format capable of 1080p50 or 60 it's the highest progressive full-HD rate.

Also, why aren't more things filmed at 1080p30 (and stored at 1080i60)? That's progressive (stored as interlaced but displayed as 30p progressive), would have more accurate motion portrayal than 1080p24 (though not as good as 1080i60) but not have the interlacing artefacts, and it should still retain the film-look. And until we have a consumer playback format capable of 1080p50 or 60 it's the highest progressive full-HD rate.
The issue is that the output needs to fit the current theaters and they can't handle 30p. Frame rate conversion is super difficult so until such time all theaters can handle 30p, I don't see the source moving in that direction.

I agree that higher frame rate progressive is desirable. I don't agree that 1080i makes the right trade off to get there :). As Tom mentioned, most of our displays are digital now so we lose a lot of vertical resolution that way. Interlace is such a nasty "non-linear" process that reversing its effect is very difficult....

Also, why aren't more things filmed at 1080p30 (and stored at 1080i60)? That's progressive (stored as interlaced but displayed as 30p progressive), would have more accurate motion portrayal than 1080p24 (though not as good as 1080i60) but not have the interlacing artefacts, and it should still retain the film-look. And until we have a consumer playback format capable of 1080p50 or 60 it's the highest progressive full-HD rate.

They would still need to convert it to the european market. 50i or 25P.

And 24P to 25P look alot better then 30P to 25P.

That is your opinion...but it will be 1080p.

I am not sure where you are seeing the overprocessing / filtering unless you are talking about some of the creative choices the artist made.

The native source looks stunning...in 32bit color...the problem is that today's encoders cannot always handle the source material we feed them. The limitation of 8 Bit Color is a huge factor in my opinion.

Part of what I will produce will be a small segment designed to make an encoder "puke" so we can see where the real limitations are with respect to current technology. I might make this particular piece of content publicly available for testing as well.

Nothing I have seen yet is truly transparent to the source ... depending on the source and how complex it is. Nature's Journey was challenging to encode....it gave the encoders and the compression tech a very hard time.


In 1080p with no overprocessed, filtered look like the last time I hope :)

Make it Au Natural ;)

Not a problem...let's coordinate and get it done..:)

Yes a D5 of Chronos would be more suitable since it would be equal for both codecs.

That is your opinion...but it will be 1080p.

I am not sure where you are seeing the overprocessing / filtering unless you are talking about some of the creative choices the artist made.

The native source looks stunning...in 32bit color...the problem is that today's encoders cannot handle the source material we feed them.

Part of what I will produce will a small segment designed to make an encoder "puke" so we can see where the real limitations are in today's technology. I might make this particular piece of content publicly available for testing as well. Nothing I have seen yet is truly transparent to the source ... depending on the source and how complex it is. Nature's Journey was challenging to encode....it gave the encoders and the compression tech a very hard time.


I understand about creative choices but personally the artists need to scale back with the digital manipulation and those distracting CGI graphics. Check out Galapagos. Now that is Nature's Journey.

Let Nature breathe.

I might even do a PIX thread for that :)

That is your opinion...but it will be 1080p.


The native source looks stunning...in 32bit color...the problem is that today's encoders cannot always handle the source material we feed them. The limitation of 8 Bit Color is a huge factor in my opinion.



Wait.

I think you are mixing pixel formats. When you say "8 Bit", that's per pixel", which is 24 bit color displayed.

What is 32bit? Are you referring to 32bit RGB format? Or were you meanign 30 or 36 bit video format?

Of course 4:2:0 chroma 4:2:0 sampling formats in the codec affect things a well, but you seemd to be talking about the source material...

Banding is in the source.

Luma levels also don't match between letterboxing and the actual image - elevated in the image area.

Yes it's true. I don't make pre-process for that. But here I want just compare grain conservation. And for this source x264 is the best by far now ...

the problem is that today's encoders cannot always handle the source material we feed them. The limitation of 8 Bit Color is a huge factor in my opinion.
That is not a problem with "encoders." There are encoders that go beyond 8 bits. The issue is that format creators in their ultimate wisdom, decided that BD (and HD DVD before it) only needed to support 8 bits. So no encoder can possibly generate more than 8 bits, lest you want the player to crash.

Nothing I have seen yet is truly transparent to the source ... depending on the source and how complex it is. Nature's Journey was challenging to encode....it gave the encoders and the compression tech a very hard time.
As was noted, this "hard time" was created by artifacts in the source requiring re-submissions. This reminds me of a little story :). We once offered Warner to encode a movie for them. Their advice? Stick to making great encoders and let us do the encoding. By the same token, I would love to see you focus your energy on making the best source you possibly can, free of banding, compression artifacts, etc. The advanced codecs are more than transparent enough to reveal issues in the source as countless threads show on AVS. To this end, they are revealing of artifacts in D-5 tape format. Little is gained from seeing how the codecs deal with pre-compressed material. That game is for another forum….

The issue is that format creators in their ultimate wisdom, decided that BD (and HD DVD before it) only needed to support 8 bits.

Thats why I hoping that any new format that make it to the market is ready from day one to go 10 or 12 bits, from the beginning.

So we dont get a 8bit solid state format. ;)

That is not a problem with "encoders." There are encoders that go beyond 8 bits. The issue is that format creators in their ultimate wisdom, decided that BD (and HD DVD before it) only needed to support 8 bits. So no encoder can possibly generate more than 8 bits, lest you want the player to crash.
Some months ago in the Industry Insiders Thread I suggested that maybe it would make sense for the encoders to accept the masters in their full bitdepth and then convert down to 8 bit internally at a late encoding stage. In my (naive) understanding of how compression works this might improve both image quality and compression efficiency because the encoder would not have to try to understand some funny dithering patterns in the source, but instead it would get to see the original smooth transients. At that time my suggestion was backed up by (IIRC) Tom McMahon and you said you would be giving it a look. I'm wondering if you ever had a chance to?

Thanks... :)

Some months ago in the Industry Insiders Thread I suggested that maybe it would make sense for the encoders to accept the masters in their full bitdepth and then convert down to 8 bit internally at a late encoding stage. In my (naive) understanding of how compression works this might improve both image quality and compression efficiency because the encoder would not have to try to understand some funny dithering patterns in the source, but instead it would get to see the original smooth transients. At that time my suggestion was backed up by (IIRC) Tom McMahon and you said you would be giving it a look. I'm wondering if you ever had a chance to?

Thanks... :)
Yeah. This was a while ago, but IIRC, the answer was something like "yes, it's interesting, but it'd take a whole lot of technical work to even valdiate the approach, so it's not something we'll get to until we tackle a lot of other low-hanging fruit for optimization." Our scaling/dithering technology is very good, which reduces the need for this somewhat.

Bear in mind there's always tons of quality tweaking and tuning to do, so there's lots of great ideas in compression that people haven't gotten around to implementing yet. Bear in mind that MPEG-2 is still getting improvements a dozen years on, and the advanced codecs have a whole lot more tools available to tweak.

Yeah. This was a while ago, but IIRC, the answer was something like "yes, it's interesting, but it'd take a whole lot of technical work to even valdiate the approach, so it's not something we'll get to until we tackle a lot of other low-hanging fruit for optimization." Our scaling/dithering technology is very good, which reduces the need for this somewhat.

Bear in mind there's always tons of quality tweaking and tuning to do, so there's lots of great ideas in compression that people haven't gotten around to implementing yet. Bear in mind that MPEG-2 is still getting improvements a dozen years on, and the advanced codecs have a whole lot more tools available to tweak.
Thanks Ben! I didn't expect that there were still any low-hanging fruits left. That's good to know. I guess that means that there may still be major improvements in store for us... :)

Thanks Ben! I didn't expect that there were still any low-hanging fruits left. That's good to know. I guess that means that there may still be major improvements in store for us... :)
Absolutely.

And when we run out of stuff to do with the current generation, H.265 is already in early development :)...

Absolutely.

And when we run out of stuff to do with the current generation, H.265 is already in early development :)...

IMO useless for high quality encoding. For example industry choose the YV12 space color because is the most efficient for quality/size. Anyway codec like H264 support already 4:4:4 encoding and lossless mode. For ultra high quality (4:4:4 in lossless mode for example) the limitation is not the codec itself but more the storage (more than 500 GB for 4:4:4 1080p lossless movie). We can't really expect than H265 will be really better than H264 for ultra high quality encoding.

IMO useless for high quality encoding. For example industry choose the YV12 space color because is the most efficient for quality/size. Anyway codec like H264 support already 4:4:4 encoding and lossless mode. For ultra high quality (4:4:4 in lossless mode for example) the limitation is not the codec itself but more the storage (more than 500 GB for 4:4:4 1080p lossless movie). We can't really expect than H265 will be really better than H264 for ultra high quality encoding.

If we were going to make another generation of codecs we should get out of the 'the best compromise format carved in stone' discussion and just support a variable bit depth and resolution independently for each luma/chroma plane, allowing it to float per GOP as needs arise. Then if either the artist, colorist, or machine encoder bit budget decided we wanted 1920x1080p 16 bit luma with only 968x320 7 bit chroma for some scene they could have it. Of course there could still be profiles for what decoders of various complexity would be required to handle.

It would not be that hard to implement these days and would probably avoid a lot of destructive pre-processing conversions.

But as a weak willed non-political software guy that is my usual recommendation when nobody can agree. Just make it an option. ;)

- Tom (not a YV12 fan)

H264 support 4:4:4 in 12 bits with lossless or lossy modes ...

H264 support 4:4:4 in 12 bits with lossless or lossy modes ...

Read this again "That is not a problem with "encoders." There are encoders that go beyond 8 bits. The issue is that format creators in their ultimate wisdom, decided that BD (and HD DVD before it) only needed to support 8 bits. So no encoder can possibly generate more than 8 bits, lest you want the player to crash."

Again, this appears to be double talk. Maybe this will explain why your communication is frustrating to me.

"That is not a problem with "encoders." There are encoders that go beyond 8 bits. The issue is that format creators in their ultimate wisdom, decided that BD (and HD DVD before it) only needed to support 8 bits. So no encoder can possibly generate more than 8 bits, lest you want the player to crash."

RJC - Read that sentence in the beginning of the paragraph and then read the sentence at the end of the paragraph. We are saying the same thing, yet you seem to say it in a way that makes me look like I was incorrect. The ecoders cannot output more than 8 bit .. for whatever reason you want to say it is ... this is the fact.

"We once offered Warner to encode a movie for them. Their advice? Stick to making great encoders and let us do the encoding. By the same token, I would love to see you focus your energy on making the best source you possibly can, free of banding, compression artifacts, etc."

RJC - This statement annoys me even more so. This is why engineers are not Producers. We do not design or "dumb down" our programming to fit within the constraints that engineers have created for us. We crate the content to be the best it can be ... for whatever reason we need ... to best tell the story or set the mood. Whether it be artistic intent using composition, angles, light, shadow, depth of field, lens focal length, and other creative aspects to filmmaking ... which appear to be lost in this form.

We do not ignore technical aspects, we are just not alwasy driven by them when it comes to what we do.

Then we massage the content as best we can to fit within the "Good Enough" box that the engineers have been so kind to design for us.

You yourself said 8 Bit COlor was an issue. That makes maybe two (2) things we have ever agreed on.

I am sorry, but we will not design content to make up for whatever limitations current technology might have. We work on a lot of content that is "evergreen" ... I want the best source content I can possibly create so that when technology takes a few leaps forward, and eventually catches up to what we are doing, I can release my stuff all over again and have content that will actually benefit from the latest advancements.

As an investor in content, it would be foolish to "Dumb Down" my content to suit today's market when I am looking at a 20 + year life on some of our products in development. I will never compromise my work to suit the limitations of current technology. We actually have some content that is 32 bit and others content that is 8K waiting for release.

Don't tell me how design programming or how to run a label ... and I won't tell you how to design a CODEC. I think that is fair...:)


That is not a problem with "encoders." There are encoders that go beyond 8 bits. The issue is that format creators in their ultimate wisdom, decided that BD (and HD DVD before it) only needed to support 8 bits. So no encoder can possibly generate more than 8 bits, lest you want the player to crash.


As was noted, this "hard time" was created by artifacts in the source requiring re-submissions. This reminds me of a little story :). We once offered Warner to encode a movie for them. Their advice? Stick to making great encoders and let us do the encoding. By the same token, I would love to see you focus your energy on making the best source you possibly can, free of banding, compression artifacts, etc. The advanced codecs are more than transparent enough to reveal issues in the source as countless threads show on AVS. To this end, they are revealing of artifacts in D-5 tape format. Little is gained from seeing how the codecs deal with pre-compressed material. That game is for another forum….

12 bit would be nice...

Thats why I hoping that any new format that make it to the market is ready from day one to go 10 or 12 bits, from the beginning.

So we dont get a 8bit solid state format. ;)

Read this again "That is not a problem with "encoders." There are encoders that go beyond 8 bits. The issue is that format creators in their ultimate wisdom, decided that BD (and HD DVD before it) only needed to support 8 bits. So no encoder can possibly generate more than 8 bits, lest you want the player to crash."

Sure. But YV12 for most user (and for industry) is the best compromise at equivalent size for perceptual quality. At equivalent size in uncompressed mode, RGB32 or YUY2 will use very lower resolution than YV12. YV12 is here for a very long time.

Hi Serguey,
I can answer that this format "AVCHD" can not work on all BD player.
I make many test with replicator, this format (AVCHD) is a BD9 with a modification of
*.bdmv file, a specific LeadIn (DVDR LeadIn and datarate at 30Mbps) and no AACS.
Many BD player which support BD9 don't play it because it don't support BD9 without AACS
and many BD player don't play BD9... (there is only Panasonic, Sony and some Pioneer BD player which play it fine).

But may be with last format "AVCRec" (BD on DVDR: Camcorder or BD+Tuner recorder),
we will have more BD player which will support it (it's like DVD-Video on CD).

He never mentioned AVCHD - so I don't know why anyone would assume that is was he was referrring to.

AVCRec has been supported by every Blu-ray player since day 1, so I would say 'all' not 'more' Blu-ray players support it.

IMO useless for high quality encoding. For example industry choose the YV12 space color because is the most efficient for quality/size. Anyway codec like H264 support already 4:4:4 encoding and lossless mode. For ultra high quality (4:4:4 in lossless mode for example) the limitation is not the codec itself but more the storage (more than 500 GB for 4:4:4 1080p lossless movie). We can't really expect than H265 will be really better than H264 for ultra high quality encoding.

Perfect for 3,900 GB HVDs (Holi-ray Discs!).

Lossless video is actually an overkill. (Not that overkills hurt either from a quality perspective)

Lossless compression is actually something you want for videoediting, were you tweak footage and dont want to start to see compressionartifacts when you do levelchange etc.

So transparent lossy 2560*1080 4:4:4 12 bit will be faster achivable, then lossless 2560*1080 4:4:4 12bit.

So heres a question for Ben, how much bandwith/storage would lossy 2560*1080P 4:4:4 12bit take?

So heres a question for Ben, how much bandwith/storage would lossy 2560*1080P 4:4:4 12bit take?
Now if we go to another encoding format, please let's also broaden the color space to handle the full color information of the film...

Now if we go to another encoding format, please let's also broaden the color space to handle the full color information of the film...
And go for more than 1080p - at least 2160p :)

Now if we go to another encoding format, please let's also broaden the color space to handle the full color information of the film...

12bit 4:4:4 is very broad. ;)

12bit 4:4:4 is very broad. ;)
No, as far as I understand it if the color space is not changed 12bit 4:4:4 will only give us finer steps between the available colors.

So heres a question for Ben, how much bandwith/storage would lossy 2560*1080P 4:4:4 12bit take?

Calcul is simple:

For 1920*1080*24 YV12
1920*1080*24*12/1024/1024 = 570 Mbps for lossless
With BD50 you compress at something like 1:20 ratio aka +/- 30 Mbps

For 4K at 24 fps in RGB32
4096*2048*24*32/1024/1024 = 6144 Mbps
With the same ratio compression at 1:20 you obtain 300 Mbps for lossy

Calcul is simple:

For 1920*1080*24 YV12
1920*1080*24*12/1024/1024 = 570 Mbps for lossless
With BD50 you compress at something like 1:20 ratio aka +/- 30 Mbps

For 4K at 24 fps in RGB32
4096*2048*24*32/1024/1024 = 6144 Mbps
With the same ratio compression at 1:20 you obtain 300 Mbps for lossy
But RGB32 is a waste of space. It stores each color (Red, Green, Blue) in 8bits. However, one whole byte (8bits) is used to store transparency information, which might have its use in computer graphics, but not in digital movie technology.

Calcul is simple:

For 1920*1080*24 YV12
1920*1080*24*12/1024/1024 = 570 Mbps for lossless
With BD50 you compress at something like 1:20 ratio aka +/- 30 Mbps

For 4K at 24 fps in RGB32
4096*2048*24*32/1024/1024 = 6144 Mbps
With the same ratio compression at 1:20 you obtain 300 Mbps for lossy
This is a layman analysis which in practice, does not hold water. And that is in addition to the valid point just made regarding not coding the alpha channel in 32-bit words.

When we encode in 8 bits from a 10 to 12 bit source, if done right, we would need to add noise (dither) to avoid banding. This dither noise actually reduces compression efficiency. So in some sense, using less bits is worse than using more! Some tests I have heard of have shown that it takes the same amount of bitrate to encode 8-bit+dither and 10-bit without dither!

On the resolution front, for most images, the pixels are more corrolated (the same) as you increase the resolution. For an extreme example, if I have a solid blue sky and I represent that in 1920 or 3840, it doesn't matter. The codec can take advantage of the similarity of the pixels and not use anywhere near the 2X the data rate to represent it. Of course, as I mentioned, this is an extreme example and in reality, the blue sky would not all be the same pixel. But similar still exist.

In addition, if a source is not computer generated, it is likely to not have a lot of high frequency information above 1920. Again, the codec can take advantage of this fact and reproduce the lower amplitude high frequency data with fewer bits than linear math would show.

Net, net, your math only works for pathological cases where the source is purely random. Real life material is not that way.

RJC - This statement annoys me even more so. This is why engineers are not Producers. We do not design or "dumb down" our programming to fit within the constraints that engineers have created for us. We crate the content to be the best it can be ... for whatever reason we need ... to best tell the story or set the mood. Whether it be artistic intent using composition, angles, light, shadow, depth of field, lens focal length, and other creative aspects to filmmaking ... which appear to be lost in this form.
Oh, you absolutely live within the engineering constraints! It's just the artists get used to the limitations of the medium.

If you're shooting 35mm film, you're not even going to try to shoot:
A live basketball game (way too much motion for 1/48th sec motion blur)

A candlelight dinner without extra lighting (no where near enough light)

Super-deep depth of field (can't do easily with 35mm lenses)

These are contraints that came out of what the engineers do. They're just old enough that everyone's used to them by now.

No, as far as I understand it if the color space is not changed 12bit 4:4:4 will only give us finer steps between the available colors.
Correct, in most implementations; black is still black and white is still white. More bits could let you define a broader gamut (since you still won't get banding if you use bigger ranges), but doesn't imply it.

Oh, you absolutely live within the engineering constraints! It's just the artists get used to the limitations of the medium.

If you're shooting 35mm film, you're not even going to try to shoot:
A live basketball game (way too much motion for 1/48th sec motion blur)

A candlelight dinner without extra lighting (no where near enough light)

Super-deep depth of field (can't do easily with 35mm lenses)

These are contraints that came out of what the engineers do. They're just old enough that everyone's used to them by now.

Definitely one of the more quotable posts here. ;)

- Tom

If you're shooting 35mm film, you're not even going to try to shoot:
A live basketball game (way too much motion for 1/48th sec motion blur)

What shutter speed do they use for football (soccer) games shot on PAL cameras? Is it 1/50th (very similar to 1/48th) of a second or higher? If it was something like 1/100th of a second wouldn't that make the motion look too choppy? So isn't the problem with film the way the 180 degree shutter works - producing gaps in motion and therefore a choppy effect, and the slow rate (not necessarily the 1/48th sec worth of motion blur)?

What shutter speed do they use for football (soccer) games shot on PAL cameras? Is it 1/50th (very similar to 1/48th) of a second or higher? If it was something like 1/100th of a second wouldn't that make the motion look too choppy? So isn't the problem with film the way the 180 degree shutter works - producing gaps in motion and therefore a choppy effect, and the slow rate (not necessarily the 1/48th sec worth of motion blur)?

It doesnt look to choppy if u have 2X "frames" to play with. Shutter effects are much easier to detect on 24fps then 50fps.

Most consumercamcorders that records in 50i or 60i uses automatic shutters, depending on the light and what the iris do. If you throw away half the fields on that video you can see how the shutters have changed.

It doesnt look to choppy if u have 2X "frames" to play with. Shutter effects are much easier to detect on 24fps then 50fps.

Most consumercamcorders that records in 50i or 60i uses automatic shutters, depending on the light and what the iris do. If you throw away half the fields on that video you can see how the shutters have changed.
Yes, but I'm sure all professional video cameras will have an option for manual shutter speeds. And wouldn't you keep the same shutter speed throughout the whole game? If shooting something like a football game on PAL video, you wouldn't want a faster shutter speed than 1/50th second unless you wanted a choppy sort of look (which didn't have the correct amount of motion blur between each field)? If there was too much light would they adjust something like the iris instead of the shutter speed?

Yes, but I'm sure all professional video cameras will have an option for manual shutter speeds. And wouldn't you keep the same shutter speed throughout the whole game? If shooting something like a football game on PAL video, you wouldn't want a faster shutter speed than 1/50th second unless you wanted a choppy sort of look (which didn't have the correct amount of motion blur between each field)? If there was too much light wouldn't they adjust the iris instead of the shutter speed?

Yes procameras does can adjust the shutters manually. I would say that its easier to leave the exposure (iris/shutters) to auto since the focus take most of your attention during fast action.

But it is no problem going over 1/50th since twice amount the frames(fields) hide the strobby look. Its a bigger problem going under 1/50th.

Yes procameras does can adjust the shutters manually. I would say that its easier to leave the exposure (iris/shutters) to auto since the focus take most of your attention during fast action.

But it is no problem going over 1/50th since twice amount the frames(fields) hide the strobby look. Its a bigger problem going under 1/50th.
I concur.

The bigger problem with higher frame rates is you get less light per frame, driving up noise in low-light.

Of course, we don't always need to use 1/2 the frame duration for shutter speed. 60p can use a 1/60th shutter speed, making for a perfect conversion to 30i if needed. But moving to 1/120th shutter could cause a problem for indoor video production. Sports often uses higher than that today, though, for outdoor stuff.

120 Hz would be a great common format (5x24, 4x30, 2x60)

I concur.

The bigger problem with higher frame rates is you get less light per frame, driving up noise in low-light.

Of course, we don't always need to use 1/2 the frame duration for shutter speed. 60p can use a 1/60th shutter speed, making for a perfect conversion to 30i if needed. But moving to 1/120th shutter could cause a problem for indoor video production. Sports often uses higher than that today, though, for outdoor stuff.

120 Hz would be a great common format (5x24, 4x30, 2x60)

I may be overlooking something but it seems with digital cameras you could use shutter speeds greater than 360, simply by averaging, as long as the shutter was always open. That should add motion blur but certainly help the low light noise problem.

- Tom

I may be overlooking something but it seems with digital cameras you could use shutter speeds greater than 360, simply by averaging, as long as the shutter was always open. That should add motion blur but certainly help the low light noise problem.

Yeah, that could work technically. Although there's really no value in capturing at >60 Hz with >360 degree shutter, since it'd get really smeary.

Maybe some kind of "Motion RAW" idea, where you grab the momentary CCD values and can "develop" that to different frame rates after the fact...

Yeah, that could work technically. Although there's really no value in capturing at >60 Hz with >360 degree shutter, since it'd get really smeary.

Maybe some kind of "Motion RAW" idea, where you grab the momentary CCD values and can "develop" that to different frame rates after the fact...

Yep. I was thinking about cameras capturing maybe a continuous 240 FPS internally but electronically averaging (in the camera) to present data at a selectable 24, 30, 40, 48, 60, 80, 120, or 240 fps with also selectable effective shutter speeds. And at the higher speeds you might want a 360 degree shutter and then apply your averaging after the fact outside the camera for slow motion effects and stuff. For instance, that would allow you to select the amount of motion blur during post if you wanted.

There wouldn't really be any shutter. Dunno if all that would work or not. It would certainly be capable of spewing out a lot of data but that is increasingly not a problem anymore.

- Tom

And at the higher speeds you might want a 360 degree shutter and then apply your averaging after the fact outside the camera for slow motion effects and stuff. For instance, that would allow you to select the amount of motion blur during post if you wanted.

At higher fps you wants higher shutter speed. So when you playback at normal speed the shutter looks like 1/48.

So 250fps need 1/500 shutters.

120fps needs 1/240 shutters.


I know they did shoot open shutters during some effects shoot in Star Wars episode 2. And then simulated the frames in between. But that only works for smoke that have a very smooth movment.

Yep. I was thinking about cameras capturing maybe a continuous 240 FPS internally but electronically averaging (in the camera) to present data at a selectable 24, 30, 40, 48, 60, 80, 120, or 240 fps...
I agree this would be a good idea too, especially for digitally shot stuff. Except for the amount of data footage at 240FPS would generate. They have 240hz motion interpolated LCDs now so it's not an impossible frame rate. And by doing such a high rate, you'd have lots of samples of the position of everything at lots of points in time to allow you to convert to lower/different rates better. So apart from the data generated (imagine 4k at 240fps) it's a good idea (though compression may make it a lot less, and there may be a good compression algorithm for such content). Maybe 120fps would be a better compromise (less data). At 4k :)

Perhaps as well as the director's themselves deciding on the end output rate (perhaps different ones for different countries), perhaps the consumers, with their players could select what rate they wanted, just like some do with motion interpolation with their TV (not player cos it's not in any yet :(). Maybe directors could decide what rate to use on a scene by scene bases - eg. talking at 24p, high speed chases at much higher rates.

At higher fps you wants higher shutter speed. So when you playback at normal speed the shutter looks like 1/48.

So 250fps need 1/500 shutters.

120fps needs 1/240 shutters.


I know they did shoot open shutters during some effects shoot in Star Wars episode 2. And then simulated the frames in between. But that only works for smoke that have a very smooth movment.

At higher fps you automatically already have higher shutter speeds, even with an open shutter. Why would it always have to be 1/2?

- Tom

The definition of a LOT OF DATA sort of keeps changing. For instance even if you kept an extravagant uncompressed 240 fps at 16 bit 4:4:4 4096x2048 that would be 4096 * 2048 * 2 * 3 * 240 bytes per second, which is about 12 GB per second or over 43 Terabytes per hour.

That seems a bit much. However, for non-long-term consumer priced drives, you can now mail order 1TB sata drives for about $135 each, so even that huge amount of data storage costs only a (rentable, reusable) $6000, not at all a show stopper compared to the costs of making a movie these days. Assume you used two drives each for redundancy and had multiple hours and you could still be under budget. And this is with quite outlandish figures.

Add a couple of huge batteries and you could ride around with it on your Seqway-cam. ;)

Though I doubt any of them can transfer that fast yet, even with fancy raid controllers. But they will soon.

- Tom

At higher fps you automatically already have higher shutter speeds, even with an open shutter. Why would it always have to be 1/2?

- Tom

You get higher shutter speeds with higher fps. But when you slowdown the movie your motionblur will not look as natural.

So 250fps with open shutter would be like 24fps with 1/250. When you do slowmotion. And you dont want strobby slowmotion.

You would like to have motionblur to look like 1/48 to mimic natural motion blur.

You get higher shutter speeds with higher fps. But when you slowdown the movie your motionblur will not look as natural.

So 250fps with open shutter would be like 24fps with 1/250. When you do slowmotion. And you dont want strobby slowmotion.
But if you had that many frames, shot open shutter, you basically have a bunch of time slices. It's be easy to blend more or fewer frames together in order to get particular motion blur scenarios.

For example, do get a 1/48th of a second shutter at 24 fps out of the 240 fps source, you'd just plend together two consecutive frames every 10 frames.

Intriguing...

The definition of a LOT OF DATA sort of keeps changing. For instance even if you kept an extravagant uncompressed 240 fps at 16 bit 4:4:4 4096x2048 that would be 4096 * 2048 * 2 * 3 * 240 bytes per second, which is about 12 GB per second or over 43 Terabytes per hour.

But if it's really a "motion RAW", we only have to store one sample per element, not three. so net bitrate may be less. Doing it fully uncompresed may be unfeasible, as the RAW CCD output tends to be quite noisy. But high quality lossy ala CineForm has shown to do well in this case.

But if you had that many frames, shot open shutter, you basically have a bunch of time slices. It's be easy to blend more or fewer frames together in order to get particular motion blur scenarios.

For example, do get a 1/48th of a second shutter at 24 fps out of the 240 fps source, you'd just plend together two consecutive frames every 10 frames.

Intriguing...

Not if your aim is slowmotion ala John Woo. ;)

Then will 1s clip ala 240fps be converted to 10s 24fps.

This is a layman analysis which in practice, does not hold water. And that is in addition to the valid point just made regarding not coding the alpha channel in 32-bit words.


It's just rapid evaluation ... but I can make more complete evaluation.


When we encode in 8 bits from a 10 to 12 bit source, if done right, we would need to add noise (dither) to avoid banding. This dither noise actually reduces compression efficiency. So in some sense, using less bits is worse than using more! Some tests I have heard of have shown that it takes the same amount of bitrate to encode 8-bit+dither and 10-bit without dither!

Dither imply more entropy and less coding efficiency. But you can imagine dither process only for usefull area (banding area).


On the resolution front, for most images, the pixels are more corrolated (the same) as you increase the resolution. For an extreme example, if I have a solid blue sky and I represent that in 1920 or 3840, it doesn't matter. The codec can take advantage of the similarity of the pixels and not use anywhere near the 2X the data rate to represent it. Of course, as I mentioned, this is an extreme example and in reality, the blue sky would not all be the same pixel. But similar still exist.

In addition, if a source is not computer generated, it is likely to not have a lot of high frequency information above 1920. Again, the codec can take advantage of this fact and reproduce the lower amplitude high frequency data with fewer bits than linear math would show.

Yes there are an empirical mathematical model for that:

C = Bitrate / ( H * W * fps ) ^ N
N in [0.50 - 0.75] interval

If you encode the same source at the same quality (aka quantizer for codec) then C will be constant.

Quality/pixel and bits/pixel


|---------------|--------------------|--------------------|--------------------|
| | files sizes | bit/(pel*fps) | bit/(pel^0.75*fps) |
|---------------|--------------------|--------------------|--------------------|
| q2 1024*576 | 3389 Kbps | 0.239 bpf | 6.63 sci |
| q2 720*400 | 1890 Kbps | 0.273 bpf | 6.33 sci |
| q2 512*288 | 1143 Kbps | 0.323 bpf | 6.32 sci |
| q2 384*208 | 714 Kbps | 0.372 bpf | 6.26 sci |
|---------------|--------------------|--------------------|--------------------|
| q3 1024*576 | 2067 Kbps | 0.146 bpf | 4.04 sci |
| q3 720*400 | 1188 Kbps | 0.172 bpf | 3.98 sci |
| q3 512*288 | 735 Kbps | 0.207 bpf | 4.07 sci |
| q3 384*208 | 462 Kbps | 0.241 bpf | 4.05 sci |
|---------------|--------------------|--------------------|--------------------|
| q4 1024*576 | 1492 Kbps | 0.105 bpf | 2.92 sci |
| q4 720*400 | 875 Kbps | 0.127 bpf | 2.93 sci |
| q4 512*288 | 548 Kbps | 0.155 bpf | 3.03 sci |
| q4 384*208 | 344 Kbps | 0.179 bpf | 3.02 sci |
|---------------|--------------------|--------------------|--------------------|
source 1280*720
resize with lanczos
encoding with XviD



|----------------|--------------------|--------------------|--------------------|
| | files sizes | bit/(pel*fps) | bit/(pel^0.75*fps) |
|----------------|--------------------|--------------------|--------------------|
| q20 1280*720 | 4026 Kbps | 0.182 bpf | 5.63 sci |
| q20 720*400 | 1728 Kbps | 0.250 bpf | 5.79 sci |
| q20 512*288 | 1057 Kbps | 0.299 bpf | 5.85 sci |
| q20 384*208 | 661 Kbps | 0.345 bpf | 5.80 sci |
|----------------|--------------------|--------------------|--------------------|
| q25 1280*720 | 2074 Kbps | 0.094 bpf | 2.90 sci |
| q25 720*400 | 924 Kbps | 0.134 bpf | 3.09 sci |
| q25 512*288 | 573 Kbps | 0.162 bpf | 3.17 sci |
| q25 384*208 | 363 Kbps | 0.189 bpf | 3.18 sci |
|----------------|--------------------|--------------------|--------------------|
| q30 1280*720 | 1023 Kbps | 0.046 bpf | 1.43 sci |
| q30 720*400 | 460 Kbps | 0.067 bpf | 1.54 sci |
| q30 512*288 | 286 Kbps | 0.081 bpf | 1.58 sci |
| q30 384*208 | 181 Kbps | 0.094 bpf | 1.59 sci |
|----------------|--------------------|--------------------|--------------------|
source 1920*1088
resize with lanczos
encoding with x264





Net, net, your math only works for pathological cases where the source is purely random. Real life material is not that way.

There are other parameters. For example in YUV mode UV chanel are really more easy to encode than Y chanel. And it's more easy with YV12 with half resolution for UV chanel. For example UV chanel for H264 is less than 10% of bit encoding.

Yes there are an empirical mathematical model for that:

C = Bitrate / ( H * W * fps ) ^ N
N in [0.50 - 0.75] interval

If you encode the same source at the same quality (aka quantizer for codec) then C will be constant.

It is certainly true that doubling the number of pixels does not double the number of bits needed to encode them in almost any real world situation.

But we have to be careful that does not become a self-fulfilling prophecy. If every thing from the lens, CCD chip, and filter chain assumes the MTF curve drops off to negligible values at DVD or lower resolution then that's what you will get. For instance there are many shows video taped from the early 80's and 90's that will never make good HDTV.

But I believe the real world actually has close to infinite resolution and we have to choose how much of that we are willing to pay for. If everyone felt we really needed cameras and image sensors that kept a mostly flat MTF curve out to 4K resolution then they would probably exist and we would suddenly find it more bit rate expensive to encode those extra details.

As unpopular as the idea is to those of us carping about DNR and low detail movies it really IS about 'good enough' and what we are willing to pay for. Nothing past filtered 1920x1080 8 bit 4:2:0 has been economically justifiable to date, and that with only a somewhat lower effective resolution after capture and encoding.

With time and money it should be possible to extend the area of the MTF curve where N in the above formula is closer to 1 (at least above .75). So we have to be careful not to assume it is some sort of universal constant.

- Tom

With time and money it should be possible to extend the area of the MTF curve where N in the above formula is closer to 1 (at least above .75). So we have to be careful not to assume it is some sort of universal constant.

- Tom

Yes with pure YUV chaotic signal. Anyway for real source N will be in [0.5 - 0.75] interval. Source with grain will be at higher N value and really clean CGI will be at lower N value.

Yes with pure YUV chaotic signal. Anyway for real source N will be in [0.5 - 0.75] interval. Source with grain will be at higher N value and really clean CGI will be at lower N value.

I prefer to think source with lots of higher frequency spatial detail will be at higher N value. Currently from telecined film that mostly means grain but I am optimistic that now we actually have consumer media that makes this (lack) visible we will develop new cameras, capture, and delivery methods that eventually deliver more of the high frequency information that is already there in the world around us.

- Tom

Not if your aim is slowmotion ala John Woo. ;)

Then will 1s clip ala 240fps be converted to 10s 24fps.
Actually, for slow-mo, you'd just grab your chunks more often. For example, for 5x slow-mo, you'd just grab every other frame without any blending. To speed up 2x, you blend together every other group of 10 frames.

Oh, I said it wrong before. For 240 to 24 frames, you'd blend FIVE frames together to match the 180 degree motion blur.

It's a whacky idea for a camera; I don't know that it would add nearly enough value to be worth the complexity.

Actually, for slow-mo, you'd just grab your chunks more often. For example, for 5x slow-mo, you'd just grab every other frame without any blending. To speed up 2x, you blend together every other group of 10 frames.


But if you extract without blending, wouldnt the motionblur be wrong as it wouldnt look like 1/48 when playback at normal speed? (24P)


It's a whacky idea for a camera; I don't know that it would add nearly enough value to be worth the complexity.

Yes it would cost more then it would gain, most filmmakers knows what type of shoot they want, and easily just setup the cam before they shoot.

So a HD-camera that can overcrank to 250fps would ne good enough for them. Better save the storage for better dynamic range.

But if you extract without blending, wouldnt the motionblur be wrong as it wouldnt look like 1/48 when playback at normal speed? (24P)
Indeed. And the blending only works if the "microframes" are 360 degree shutter.

Wow, no Codec Wars posts for a month! That can't be allowed to stand :).

In the interest of peace and unification, Microsoft's Gary Sullivan was recently on hand along with other representatives of the Joint Video Team to recieve a Technical Emmy for H.264 High Profile.

http://on10.net/blogs/benwagg/Technical-Emmy-for-H264MPEG-4-AVC/

Just to remind everyone that Microsoft has been quite involved with H.264 as well as VC-1.

Wow, no Codec Wars posts for a month! That can't be allowed to stand :).

In the interest of peace and unification, Microsoft's Gary Sullivan was recently on hand along with other representatives of the Joint Video Team to recieve a Technical Emmy for H.264 High Profile.

http://on10.net/blogs/benwagg/Technical-Emmy-for-H264MPEG-4-AVC/

Just to remind everyone that Microsoft has been quite involved with H.264 as well as VC-1.

And here is an excellent presentation by Gary J. Sullivan, Pankaj Topiwala, and Ajay Luthra on the topic back in 2004:

http://www.fastvdo.com/spie04/spie04-h264OverviewPaper.pdf

Wow, no Codec Wars posts for a month! That can't be allowed to stand :).

In the interest of peace and unification, Microsoft's Gary Sullivan was recently on hand along with other representatives of the Joint Video Team to recieve a Technical Emmy for H.264 High Profile.

http://on10.net/blogs/benwagg/Technical-Emmy-for-H264MPEG-4-AVC/

Just to remind everyone that Microsoft has been quite involved with H.264 as well as VC-1.

Dear Ben,

you are asking for Codec Wars and you get it...;)

Universal (this tiny little studio which has such an close relationship with MS and even "advertised" VC-1 on each of their latest back covers) switched to AVC encoding with their latest high profile release - The Incredible Hulk.

If this continues with Hellboy II and Wanted later this year and Universal uses AVC from now on the industry practically decided this "Codec War" - wouldn't you agree? Warner being the last holdout...

Dear Ben,

you are asking for Codec Wars and you get it...;)

Universal (this tiny little studio which has such an close relationship with MS and even "advertised" VC-1 on each of their latest back covers) switched to AVC encoding with their latest high profile release - The Incredible Hulk.

If this continues with Hellboy II and Wanted later this year and Universal uses AVC from now on the industry practically decided this "Codec War" - wouldn't you agree? Warner being the last holdout...

Maybe WB is using VC1 with TDK because its the only codec powerful enough to handle the switches in aspect ratio without becoming blurry or blocking?

Maybe WB is using VC1 with TDK because its the only codec powerful enough to handle the switches in aspect ratio without becoming blurry or blocking?

That makes absolutely no sense whatsoever.

That makes absolutely no sense whatsoever.
Amen.

Both are very capable codecs. Their choice of VC-1 could merely be a workflow efficiency decision.

Amen.

Both are very capable codecs. Their choice of VC-1 could merely be a workflow efficiency decision.

This is also not valid anymore (maybe this is the reason for WB to switch to AVC). New Thomson and Sony encoders are scalable, so you can get twice faster than real time encoding if you realy need.

Andrew

This is also not valid anymore (maybe this is the reason for WB to switch to AVC). New Thomson and Sony encoders are scalable, so you can get twice faster than real time encoding if you realy need.

Andrew
Don't know about Thomson but Sony encoders (as with Microsoft's VC-1) have always been scalable, running on multi-CPU PCs.

And there is a lot more to workflow than encoding speed....

so... anybody still uses VC-1 ?

so... anybody still uses VC-1 ?

"How The West Was Won" - Warner Brothers

okay, allow me to re-iterate: is there any new releases planned will be using VC-1?

okay, allow me to re-iterate: is there any new releases planned will be using VC-1?
What would you like the answer to be?

I want the answer to be "yes". That's why I asked :D I like VC-1 encodes from the HD DVD days.

I want the answer to be "yes". That's why I asked :D I like VC-1 encodes from the HD DVD days.
"You are a good man Charlie Brown :)".

"How The West Was Won" - Warner Brothers

These are Xylon's opinions, and I was also really impressed with this BD.

"In terms of overall picture quality compared to other restored catalog titles, this transfer tops them all. Better than Grand Prix HD DVD or The Searchers. It blows them away. Tired of reading posts like "Old movies can never look as good as the blockbuster movies of today!!!" ? I will tell you that this 45 year old movie is sharper, clearer and more detailed than most *ahem* TIER 0 PQ Blu-rays or HD DVDs.

Filmed using Cinerama process it is just f****** gorgeous. Watching the movie using the "Smilebox" version is truly an experience. The depth of field is amazing giving you the effect of "3D" or "looking thru a window". I still remember renting the laserdisc version from Dave's Video at Studio City many years ago in glorious widescreen 480i and digital stereo sound I don't want to sound elitist but after watching this movie and some of its scenes multiple times on different viewing sets of mine(check my profile for a list of my HT equipment), anything less than a 1080p projector is not recommended (my opinion). Now if you have a curved screen then its even better

No noticeable EE observed. Mild DNR is present on some scenes. But not to the level of Patton or Longest Day. Film grain is intact. DNR dial is not cranked up to "11" like those two. No waxy faces here. Fox studios should contact WB on the proper use of DNR on HD catalog titles."

VC1 rocks... :)

so... anybody still uses VC-1 ?
I think most WB releases are still in VC-1. Weinstein's new Diary of the Dead was also VC-1, so perhaps they will continue to use VC-1 rather than AVC. I think Universal may have moved over to AVC since Incredible Hulk uses AVC rather than VC-1.

Band of Brothers in high bitrate VC-1 is beautiful with lots of grain fully intact.:)

I think most WB releases are still in VC-1. Weinstein's new Diary of the Dead was also VC-1, so perhaps they will continue to use VC-1 rather than AVC. I think Universal may have moved over to AVC since Incredible Hulk uses AVC rather than VC-1.

Band of Brothers in high bitrate VC-1 is beautiful with lots of grain fully intact.:)

Dreamworks seems to use VC1 on some movies.

Band of Brothers in high bitrate VC-1 is beautiful with lots of grain fully intact.

Not as much as was removed. ;) HBO is on an anti-grain crusade.