CoreCodec/H.264 Codec "CoreAVC"

[Hans Ohlo]

So when the levels conversion checkbox is on, levels are left as is, and when it's off they are further compressed to TV range, that gives 32-226 for your samples.
why do that? a decoder is (at default configuration) supposed to decode what is present in the files. so why compress the colorspace when a conversion flag is _not_ set? even at tv levels form 16-235 there can be blacker-than-black or whiter-than-white levels below 16 oder above 235. why compress these to the standard tv levels. on a properly calibrated ouput such a thing will look totally wrong!

[nk]

I install CoreAVC on Windows Vista Ultimate 32bit.
installing processing seems good, but when I configure,
an error dialog(RunDLL) appears and says initializing routine failed
or so. (Sorry my windows is not english version so, cannot tell exact error message.)

How to fix my problem?

[foxyshadis]

why do that? a decoder is (at default configuration) supposed to decode what is present in the files. so why compress the colorspace when a conversion flag is _not_ set? even at tv levels form 16-235 there can be blacker-than-black or whiter-than-white levels below 16 oder above 235. why compress these to the standard tv levels. on a properly calibrated ouput such a thing will look totally wrong!

Then the fullrange should be off. You're not thinking this through, so you're directing your vemon a Core for properly supporting an MPEG feature; complain to Sky. You might need to grab a hex editor or write a script to switch these VUI entries off.

Here's a table of how this all works, listed by input & output:

fullrange
on off
----+------------
| leave expand
C PC| 0-255 16-235
o | 0-255 0-255
r |contract leave
e TV| 0-255 16-235
| 16-235 16-235

What Core calls "fix VMR9 range" means turn what I call PC levels on, unchecking it is what I call TV levels. Whether it needs to be expanded or compressed or left alone depends on what gets fed into it, and the stream should signal what's in it - of course it's shouldn't signal incorrectly, as in this case. Otherwise fullrange is assumed off, and since most streams don't have that data, you only experience the right-hand side of the table.

Fullrange on + TV mode means all luma gets compressed. Since the reality is that the stream's fullrange is off, luma is wrong, and since you're outputting YUV, I think, it compounds when the display then assumes fullrange YUV->RGB conversion, even though the stream isn't, because video driver writers keep breaking it. Nice little comedy of errors there.

There appears to be a DXVA flag that can be sent to the video renderer to specify fullrange on or off, described here (http://msdn2.microsoft.com/en-us/library/ms796493.aspx). My gut feeling is that as usual, cards will ignore it, but it's worth trying and complaining to driver writers if they get it wrong. Then at least the problem wouldn't be compounded. If this was used, Core could add a "default" option to the range conversion, which would simply pass it on to the card unchanged whether fullrange was on or off.

[BetaBoy]

Core could add a "default" option to the range conversion, which would simply pass it on to the card unchanged whether fullrange was on or off.

Very well stated... we already added this option for the next release. Also we noted that the 'VMR9 Fix' name might need to change.

[BlackSun]

I install CoreAVC on Windows Vista Ultimate 32bit.
installing processing seems good, but when I configure,
an error dialog(RunDLL) appears and says initializing routine failed
or so. (Sorry my windows is not english version so, cannot tell exact error message.)

How to fix my problem?

Try to re-install it, but click the details button at the end of the installation and see if the registration of the filter went well.

[nk]

Try to re-install it, but click the details button at the end of the installation and see if the registration of the filter went well.

Output folder: C:\Program Files\CoreCodec\CoreAVC Professional
Extract: CoreAVCDecoder.ax... 100%
Extract: CoreAVCReadme.txt... 100%
Could not load: C:\Program Files\CoreCodec\CoreAVC Professional\CoreAVCDecoder.ax
Could not load: C:\Program Files\CoreCodec\CoreAVC Professional\CoreAVCDecoder.ax
Created uninstaller: C:\Program Files\CoreCodec\CoreAVC Professional\CoreAVC Professional-uninstall.exe
Create folder: C:\Users\nkoriyama\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\CoreCodec\CoreAVC Professional
Create shortcut: C:\Users\nkoriyama\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\CoreCodec\CoreAVC Professional\CoreAVC Readme.lnk
Create shortcut: C:\Users\nkoriyama\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\CoreCodec\CoreAVC Professional\Configure CoreAVC.lnk
Create shortcut: C:\Users\nkoriyama\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\CoreCodec\CoreAVC Professional\Re-Register CoreAVC.lnk
Create shortcut: C:\Users\nkoriyama\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\CoreCodec\CoreAVC Professional\Un-Register CoreAVC.lnk
Completed


This is a installation log of CoreAVC 1.3.0, and not registered as directshow filter.
I have two licenses of CoreAVC, one is used on XP works well,
but another one on Vista does not work at all.

[DeepBeepMeep]

Very well stated... we already added this option for the next release. Also we noted that the 'VMR9 Fix' name might need to change.

An option to force CoreAvc to keep the color levels as they are in the stream (and ignore all the flags) is exactly what we need here. I hope however that your are not referring to the early 1.3 version option which has replaced the 'VMR9 Fix' (that is "Convert TV scale to PC scale"). Although this option is potentially useful it didn't solve the issue.

[Hans Ohlo]

Very well stated... we already added this option for the next release. Also we noted that the 'VMR9 Fix' name might need to change.
oh thanks.

and please add clear descriptions in the readme, so that it is clear what these options manipulate and what not. i can not live with multiple color space convertions which everytime degrade color precision.

OH and... PLEASE get the next release with these fixes sooner than 4-6months from now. minor service releases like 1.3.0.2 or so are very welcome (not only be me i think ;)).

-hans

[DigitalDeviant]

I'm having the same filter registration problem as nk, the error given when trying to unregister, reregister or change the configuration is:

Error Loading C:\....

A dynamic link library (DLL) initialization routine failed.

I'm on XP (logged in as the administrator). I'm already working on this via a trouble ticket but figured it wouldn't hurt to post here too. One (not-so-)small complaint is that it won't let me try to reinstall 1.2. I don't see any reason not to be able to roll back to a previous version.

[nk]

I found a reason why dll can't load, because my system has something wrong with ieframe.dll and shlwapi.dll. I found it by dependency walker.

[DigitalDeviant]

I found a reason why dll can't load, because my system has something wrong with ieframe.dll and shlwapi.dll. I found it by dependency walker.

shlwapi.dll almost always seems to show something wrong, there is a function that doesn't work or something like that which I don't recall off hand now. Maybe ieframe.dll is the problem on yours but I tried the same thing on my end and I have no problems with it. I found a few missing dll's (dwmapi.dll & efsadu.dll) but neither of those was the culprit.

Anyway, I haven't heard back from core support since they emailed me a download link that only pulls up an error page after I log in.

[ACrowley]

For you all... Also we are going to be doing a new CoreAVC release tomorrow for 1.3.0.1 as we have found a serious bug in 1.3 (thanx for the reports everyone!)

is 1.3.0.0 updated now ? cant see a fixed 1.3.0.1 ?

Can only download the 1.3.0.0 from my Account

postponed again ?

[BetaBoy]

I noted we are fixing the reported issues and that we are adding the additional features...

Here are screen caps of the new options...


http://img504.imageshack.us/img504/7167/clipboard01dl4.jpg (http://imageshack.us)

http://img524.imageshack.us/img524/3152/clipboard02vu8.jpg (http://imageshack.us)

[BetaBoy]

Anyway, I haven't heard back from core support since they emailed me a download link that only pulls up an error page after I log in.
The staff will be back on Monday.... atm its only a few ppl that are having the issue and we are trying to see what the common issue is for them but have not found it as of yet.

[foxyshadis]

I'm curious, but can you explain what the difference between aggressive and normal deinterlacing is? Do any of the new options do IVTC instead of deinterlacing?

[Hans Ohlo]

Here are screen caps of the new options...

http://img504.imageshack.us/img504/7167/clipboard01dl4.jpg
http://img524.imageshack.us/img524/3152/clipboard02vu8.jpg

ok very nice ;),
will the levels in the stream be left untouched if i select in:pc out:pc?

[DigitalDeviant]

The staff will be back on Monday.... atm its only a few ppl that are having the issue and we are trying to see what the common issue is for them but have not found it as of yet.

So no one is working on this until Monday and even then it's anyones guess when I can get a working decoder? I wouldn't be so upset if I could reinstall 1.2, which worked, but the installer prevents this.

[bob0r]

I'm curious, but can you explain what the difference between aggressive and normal deinterlacing is? Do any of the new options do IVTC instead of deinterlacing?

If i am correct, when we reported H.264 MBAFF streams did not always deinterlace, was because certain features were not detected in the stream, it should be in the forum somewhere, there were like 4 or 5 points where the decoder looks for interlacing.
Now when all those fail, but you are certain you need deinterlacing, you enable the agressive mode, and it will always deinterlace your stream.

[nk]

Betaboy, please answer.

http://forum.doom9.org/showthread.php?p=974113#post974113
http://forum.doom9.org/showthread.php?p=974408#post974408
HOW CAN I RUN CoreAVC on Windows Vista?
WHAT IS WRONG? MY ENVIRONMENT? or Windows Vista? or CoreAVC?

[eLupus]

vista is missing the product key for windows that coreavc's installers use to validate your install.

[BetaBoy]

nk.... visit http://support.corecodec.com and submit a ticket and our team will walk you through whats going on.

[BetaBoy]

As requested... here is the CoreAVC 1.3.x Options descriptions... we are adding Tooltips to the installer as well. Please comment on anything that may need clarification.

Accept Input
This setting controls which DirectShow Media Types the decoder accepts on input. Uncheck only if you are troubleshooting problems with CoreAVC incorrectly decoding some variant of H.264, or want to use another decoder for it.

Deblocking
This setting controls how the deblocking step of H.264 specification is executed by the decoder. Deblocking is a complex process that consumes significant processing resources. If your machine is not fast enough, you might want to turn off deblocking for some frames, but it will degrade visual quality.

Standard deblocking - do deblocking exactly as specified by H.264
Skip b-frame deblocking - skip deblocking step when decoding B-frames.
Skip all deblocking - do not perform deblocking at all.

Deinterlacing
This setting specifies how interlaced material is handled by the decoder.

Software deinterlacing - the decoder outputs one deinterlaced frame for each pair of fields.
Software double framerate - the decoder outputs one deinterlaced frame for each input field.
DirectShow deinterlacing - the decoder does not deinterlace video, but flags it as interlaced where appropriate, so Video Renderer filter can deinterlace it.
No deinterlacing - the decoder does not deinterlace video, and flags it as progressive on output.

Aggressive deinterlacing
This option determines how the decoder detects interlacing in source stream.

Off - use only picture timing SEI and POC numbers to detect interlaced video. Unfortunately not all encoders properly flag material as interlaced.
On - in addition to SEI and POC, assumes source is interlaced if any interlaced coding tools are used in encoding the frame (MBAFF, PAFF).

Crop 1088 to 1080
H.264 encoded video size is always a multiple of 16, and sequences that are 1080 pixels high are encoded as 1088 padded at the bottom. Also H.264 specifications provides a set of cropping parameters to signal that parts of the encoded picture are not important and should not be displayed. Some H.264 encoders fail to specify cropping parameters when encoding 1080 video.

Off - do not crop video.
On - when input video is exactly 1088 pixels high, crop 8 pixels off the bottom.

Force VMR AR correction
This option can be used if you are working with the decoder outside the normal player environment.

Off - does not change VMR settings.
On - instructs VMR filter to maintain aspect ration of the video that it displays. Normally this AR correction is the responsibility of a video player. This option should normally be off.

Color space output
This setting determines the preferred output color space. The decoder tries each format in order specified until it is accepted by the Video Renderer filter.

Levels
H.264 specification allows two variants of Luminance range encoding. The default one, so called TV levels, sets the black level at 16 and white level at 235, and allows blacker-than-black and whiter-than-white levels. Another option is full-range luminance, so called PC levels, it has black at 0, and white at 255. H.264 streams by default use TV levels, but a full-range flag may be specified that signals PC levels use in the stream. Change these options only when you are troubleshooting incorrectly encoded streams, or use a misbehaving Video Renderer.

Input levels
TV (16-235) - always assume the stream uses TV levels
PC (0-255) - always assume the stream uses PC levels
Autodetect - use the fullrange flag in the stream to determine Luminance range

Output levels, this also affects conversion to RGB color space when it is done by the decoder.
TV (16-235) - assume the Video Renderer expects TV levels
PC (0-255) - assume the Video Renderer expects PC levels
Autodetect - use PC levels when VMR is used as a Video Renderer, and TV levels for all others.

[dchard]

BetaBoy: I'am sorry if someone asked before, but when th GPU support will be added? With the upcoming 1.3 version or later?

Thanks!

Dchard

[ChronoCross]

it's been asked about 800 times. And the answer is the same....

"No comment at this time"

[clsid]

Meanwhile all early adopters have already been waiting for almost exactly 1 year (!!!) for the promised GPU support :eek:

I say class action lawsuit :devil:

[ChronoCross]

Meanwhile all early adopters have already been waiting for almost exactly 1 year (!!!) for the promised GPU support :eek:

I say class action lawsuit :devil:

for a feature you knew wasn't in the product at purchase and no definitive date was set? That'll get about 5 seconds into it before the judge pounds the gavel and says dismissed.

[tomos]

for a feature you knew wasn't in the product at purchase and no definitive date was set? That'll get about 5 seconds into it before the judge pounds the gavel and says dismissed.

so does that mean you can add any 'feature' you want?

*guaranteed to use less than 1% CPU time

* not a guarantee! feature to be added in the future at some random time

so is that how nvidia/ati always get away with their various *'claims'?

*lies

[clsid]

Consumer law says that if a product doesn't deliver what can reasonably be expected of it, then a consumer is entitled to a refund.

Any judge will agree with me that 1 year is not a reasonable delay considering how the feature was marketed here on Doom9.


/me feels sorry for everyone who bought CoreAVC for its GPU support.

[Inventive Software]

/me waits till CoreCodec actually make it open-source before critisising it's lack of GPU, since somebody else could probably add it before they do! :D

[clsid]

Adding GPU support is probably much much more difficult than they anticipated. They are also dependent on ATI/nVidia for getting the required information. But CoreCodec should be more honest about things and just say it takes a couple more months.

[Zag]

Just curious, but why is GPU support so important now days? With dual-core cpu pretty much mainstream why even bother.

[ChronoCross]

Consumer law says that if a product doesn't deliver what can reasonably be expected of it, then a consumer is entitled to a refund.

Any judge will agree with me that 1 year is not a reasonable delay considering how the feature was marketed here on Doom9.


/me feels sorry for everyone who bought CoreAVC for its GPU support.

glad your not my lawyer. Consumer law says nothing about features that will be added. There is no such thing as a reasonable time frame for something that never had a timeframe in the first place. As long as it's added in the 1.x generation then they have followed all laws.

Adding GPU support is probably much much more difficult than they anticipated. They are also dependent on ATI/nVidia for getting the required information. But CoreCodec should be more honest about things and just say it takes a couple more months.

They already said it's going to take more time. They have no reason to tell you the reasoning and are likely under some agreement to that effect. Even if they aren't it's simply a marketing decision. It's still the most efficient software decoder around.

Just curious, but why is GPU support so important now days? With dual-core cpu pretty much mainstream why even bother.

Agreed. I have a single core 2.4 ghz athlon 64 and can play 1088i mbaff with AC3 audio. Dual core would make it a cinch.

[tomos]

bacause it's a cpu that is unused? if you want to leave something encoding in the background, it's kind of a waste of time if just watching a movie at the same time takes up a large amount of cpu time.

[clsid]

glad your not my lawyer. Consumer law says nothing about features that will be added. There is no such thing as a reasonable time frame for something that never had a timeframe in the first place. As long as it's added in the 1.x generation then they have followed all laws.

GPU support *IS* listed as a feature on their website. People buy the product with that in mind. Compare it with buying a car and a spare wheel that has an unknown delivery delay.

According to you they can add it 10 years from now since they didn't specify a date. Or even a day after the goddamn sun has burned out. I say that time DOES matter, no matter if a specific date has been given or not. It should be delivered within a *REASONABLE* timeframe. Specially since they raised the expectation that it could be expected shortly after the 1.0 release. If they fail to deliver the full product as it was advertised within a timely fashion, then users should be entitled to a refund.
They already said it's going to take more time. They have no reason to tell you the reasoning and are likely under some agreement to that effect. Even if they aren't it's simply a marketing decision.
You obviously don't know their reasoning. So don't use your guesses as an argument. Perhaps they are failing to make it work properly.
It's still the most efficient software decoder around.
That's a fallacy.

[pookguy88]

Is there Vista support yet?
1.3 available yet?

[tomos]

some versions work fine in vista. dont know about the official ones though

[ChronoCross]

GPU support *IS* listed as a feature on their website. People buy the product with that in mind. Compare it with buying a car and a spare wheel that has an unknown delivery delay.

According to you they can add it 10 years from now since they didn't specify a date. Or even a day after the goddamn sun has burned out. I say that time DOES matter, no matter if a specific date has been given or not. It should be delivered within a *REASONABLE* timeframe. Specially since they raised the expectation that it could be expected shortly after the 1.0 release. If they fail to deliver the full product as it was advertised within a timely fashion, then users should be entitled to a refund.

You obviously don't know their reasoning. So don't use your guesses as an argument. Perhaps they are failing to make it work properly.

That's a fallacy.

It's listed but says "added at a later date" perhaps you missed that. It also doesn't specify a date. If you bought it for GPU support you should have waited till it was implemented and you lose the right to complain that it;s not ready.

Yes they can add it in 10 years if they wanted. As long as it's released within the 1.x branch. I think the sun burns out is a bit overkill as all purchasers of coreavc will be long dead.

As for faster software decoder I think your misplacing people who already support GPU with coreavc pure software decoding. I haven't yet seen a legitimate non-biased test showing that anything is better. ffdshow is for sure not faster and doesn't support the features nor do I expect it to for a free software project. I accept it's limitations.

[clsid]

I am not blind. I know exactly what it says on the website. That is also my point. They shouldn't list it as a (future) feature if they can't deliver it in a reasonable time.

Again, the fact that CoreAVC performance is already good without GPU support is of no relevance to this discussion.

[Weltall]

I agree with you clsid :)

Btw, what about that GPU support? If there is it, cpu will not be used, just graphics card? 0% cpu use to decode video???? +__+

Could someone explain me what everyone is complaining about this level and color stuff in coreavc? I didn't get well. Is coreavc showing "pc level" videos like "tv level"'s in some unusual case, like an option enabled which had to be enabled, but couldn't do this kind of problem? I'm stucked :P

[_xxl]

Please try nero's avc decoder which is multithreaded and DXVA and see how much cpu it uses.

[JohnnyFu]

Btw, what about that GPU support? If there is it, cpu will not be used, just graphics card? 0% cpu use to decode video????

The GPU takes over specific arithmetic operations, this can reduce the CPU usage significant. On some files it decreases the CPU usage from 90-100% to 50-60% on my single core AMD K8 (cyberlink decoder).
But there still remains alot of work for the CPU.

[Jay Bee]

Any news on the release date for a CoreAVC version with the AVI bug fixed?

And concerning GPU support it's worth mentioning that I have yet to see a GPU assisted AVC decoder that performs deblocking. This is probably ATI/Nvidia's responsibility so IMO it's unlikely that a CoreAVC GPU version would be any different.

[Disabled]

And concerning GPU support it's worth mentioning that I have yet to see a GPU assisted AVC decoder that performs deblocking. This is probably ATI/Nvidia's responsibility so IMO it's unlikely that a CoreAVC GPU version would be any different.

Unfortunately they have to! They advertised a GPU accelerated H264 Decoder and deblocking is part of the standard. If they won't allow deblocking combined with GPU accel, they have a sub H264 implementation GPU accelerated and thats not what they advertised.

[Jay Bee]

I agree but ATI/Nvidia/Cyberlink etc. have been getting away with it for quite a while now. In fact I have never even seen this issue mentioned on any of the internet computing sites.

[clsid]

Maybe DXVA2 is better in that aspect.

[HookedOnTV]

I've got jerky playback. Running Vista with CoreAVC v1.3, x2 3800. Through the process of elimination I found the 720p encodes(x264, 1280x720, 4500kbs, HQ-Slow) play back jerky. Same clip at 544p plays smooth. CPU utilization never goes above 40% on the 720p clip so I am lost at this point. Is there something wrong with CoreAVC?

[ChronoCross]

I've got jerky playback. Running Vista with CoreAVC v1.3, x2 3800. Through the process of elimination I found the 720p encodes(x264, 1280x720, 4500kbs, HQ-Slow) play back jerky. Same clip at 544p plays smooth. CPU utilization never goes above 40% on the 720p clip so I am lost at this point. Is there something wrong with CoreAVC?

I'd say there is a problem with something on your machine. Are you using IDE drives or sub par memory? there has got to be a bottleneck with your system hardware.

[Jay Bee]

I've got jerky playback. Running Vista with CoreAVC v1.3, x2 3800. Through the process of elimination I found the 720p encodes(x264, 1280x720, 4500kbs, HQ-Slow) play back jerky. Same clip at 544p plays smooth. CPU utilization never goes above 40% on the 720p clip so I am lost at this point. Is there something wrong with CoreAVC?

Are the files 60 fps? Do they work with Xvid? Do you have an AGP card? I had similar problems when I used AGP cards but they've gone after switching to PCIE.

[HookedOnTV]

23.976 fps. IDE hard drive but the files are coming across gbit. Memory is Crucial DDR400 running 1:1. PCIe Gforce 7600GT. All other formats play fine (xvid, wmv, mpeg2), although h264 stuff is going to be the hardest. Recent Vista install. Ran all the updates. Got the latest drivers. Installed ffdshow, CoreAVC and Haali Media Splitter.

[Ronin-7]

23.976 fps. IDE hard drive but the files are coming across gbit. Memory is Crucial DDR400 running 1:1. PCIe Gforce 7600GT. All other formats play fine (xvid, wmv, mpeg2), although h264 stuff is going to be the hardest. Recent Vista install. Ran all the updates. Got the latest drivers. Installed ffdshow, CoreAVC and Haali Media Splitter.

You don't have Cool & Quiet enabled in the bios by any chance ?