View Full Version : CoreCodec/H.264 Codec "CoreAVC"
Jay Bee
19th March 2007, 09:55
I tried out those samples using WMP and MPC and they play fine. I would imagine this is a DVBViewer incompatibility with coreavc and not coreavc itself.
Argh, you made me waste another hour of my life taking a closer look at those samples. :)
What I quickly found out is that the decoding problems only occur when using HDTVPump instead of Haali. Yet Haali Splitter causes problems of its own: With the samples that have MP2 audio it signals the sub-type on the audio pin as {00000055-0000-0010-8000-00AA00389B71}. This is the same sub-type that is used by MP3. Depending which player you are using and which audio decoders you have installed this can cause playback to freeze. MPC can deal with it, WMP can deal with it if the right codecs are installed and ZP can only deal with it by setting up some advanced filter rules.
While I was at it I also checked the deinterlacing on those samples. The only file that didn't deinterlace although it clearly contains interlaced content was "anixe.hd.ts" while the 5 other files that are true interlaced all worked fine. So I'm guessing that's not CoreAVC's fault so overall interlaced playback works great with 1.3.
oddball
19th March 2007, 22:08
What new 'features' does 1.3 have? GPU support? If not I'll pass and continue using the much better (and free) ffdshow even though I am registered for the free upgrade. Burned once. Never again.
BetaBoy
20th March 2007, 00:33
What new 'features' does 1.3 have? GPU support? If not I'll pass and continue using the much better (and free) ffdshow even though I am registered for the free upgrade. Burned once. Never again.
oddball... not to get into the tit-for-tat... but i'd be interested in specifics of what your opinion is that determines 'better'. Like faster? More options? You can PM it to me if you don't wanna go to OT in the thread.
On GPU.... not to the public yet.
wiak
20th March 2007, 00:34
-playback of AVC in AVI is completely garbled
that should be garbage by default, also avi is a crappy old container that never was made for h264, REMUX it at once!
:helpful:
BetaBoy
20th March 2007, 00:41
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!)
mrkazador
20th March 2007, 01:04
There still is an issue with coreavc creating artifacts with encodes.
http://forum.doom9.org/showthread.php?p=945658#post945658
Im not sure if this will be fixed........
BetaBoy
20th March 2007, 03:01
If you have a sample video PM me and i'll give you a dropbox to upload the video to so we can look at it.
mrkazador
20th March 2007, 06:21
If you have a sample video PM me and i'll give you a dropbox to upload the video to so we can look at it.
Pm sent
Anyone want to do their own testing, you can download this sample clip.
http://rs38.rapidshare.com/files/21985628/Doom9.rar
Notice the person on the left @ 00:05, his collar shows some serious artifacts. Again at 00:10....
This also happens with cyberlinks decoder but NOT with ffdshow.
Jay Bee
20th March 2007, 06:30
that should be garbage by default, also avi is a crappy old container that never was made for h264, REMUX it at once!
:helpful:
Hehe, why can't doom9 people attach their emotional support to football teams or racing drivers or something, like normal people. Believe me it's more fun than being a container format fanboi (or basher). :)
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 there a way to tell from the download page if the available download is a new version or not?
BetaBoy
20th March 2007, 10:31
Jay Bee .... i'll add a custom field for version/date there so you can see the difference.
Maxiuca
20th March 2007, 14:09
BetaBoy, can you finally reveal what is new in 1.3? Cause all we can get from you that there is a new version out, like renumbering the version was itself a priority. But what is new? When there's gonna be GPU support?
_xxl
20th March 2007, 15:19
Do we really need DXVA?CoreAVC is multithreaded and working ok with H.264 HD-DVD on X2.
ToS_Maverick
20th March 2007, 15:21
i can barely play 1080p files on my laptop, with DXVA, i think, it would be possible!
Theliel
20th March 2007, 15:26
ppl read somethings?
betaboy: On GPU.... not to the public yet.
so GPU support will be release when are will be ready.
needed??? well, nothing in this world are needed, but have hardware support for h264 decoder are very nice feature. In this moment, for example, I use Cyberlink H264/AVC decoder because its the only one decoder with Hard support that work for me. If CoreAVC add hard support too, I sure this help to much ppl.
less CPU load more ppl can play 1080p h264 or more sofisticated decoder can be build ;)
AKarpo
20th March 2007, 15:45
ppl read somethings?
betaboy: On GPU.... not to the public yet.
so GPU support will be release when are will be ready.
needed??? well, nothing in this world are needed, but have hardware support for h264 decoder are very nice feature. In this moment, for example, I use Cyberlink H264/AVC decoder because its the only one decoder with Hard support that work for me. If CoreAVC add hard support too, I sure this help to much ppl.
less CPU load more ppl can play 1080p h264 or more sofisticated decoder can be build ;)
Question, how do you use Cyberlink's H264 codec properly to have hardware accelerated .mkv playback? I'm using Media Player Classic, and I can't figure it out. One thing I am worried about is that my old 6800GT won't support MPEG-4 accelerated playback (Purevideo says 6800GT AGP's will not do accelerated MPEG-4), but maybe it will work if it's done solely through DXVA? Feel free to pm me so we don't get yelled at for hijacking the thread.
Anyways, despite the gnashing of teeth of many users on this forum, I very much like CoreAVC and can't wait for GPU support to put my newly purchased 8800GTS to work. All those transistors doing hardware accelerated 1080p...Be still my beating heart!
One last note, BetaBoy, any ETA for 1.3.0.1 dropping sometime soon?
_xxl
20th March 2007, 15:50
How do you enable DXVA for ASUS 7900 TOP?
foxyshadis
20th March 2007, 16:15
(Purevideo says 6800GT AGP's will not do accelerated MPEG-4)
That's right, that card's hardware is just plain broken, not something that can be worked around in software. There are hacked drivers somewhere that turn DXVA on for it, and it just returns garbage. It was the very first 6-series card, first attempt at purevideo, and they blew it, sorry.
Theliel
20th March 2007, 16:17
this is offtopic, I send us now a PM to both :)
dentdiz81
20th March 2007, 19:51
'less CPU load more ppl can play 1080p h264 or more sofisticated decoder can be build '
If there was a way to get core avc to work with mplayer (in windows and linux) 1080p would be a breeze for virtually all computers because it is so lightweight. It is not so much the codec but the players they are used in.
Hey betaboy, if you how to make this possible, or know of a player as light as mplayer, let me know. Hopefully someone has cracked it.
By the way i have tried TCMP, kmplayer,bsplayer etc .
mrkazador
20th March 2007, 20:47
'less CPU load more ppl can play 1080p h264 or more sofisticated decoder can be build '
If there was a way to get core avc to work with mplayer (in windows and linux) 1080p would be a breeze for virtually all computers because it is so lightweight. It is not so much the codec but the players they are used in.
Have you tried Zoom Player Pro? A combination of codecs should allow for perfect playback. I dont think mplayer is any different than any other media player like zoom player. Unless you can prove me wrong....
SealTooGreat
20th March 2007, 21:32
^MPlayer is different 'cause it use its own codecs by default, but you can force it to use DirectShow for example.
If there was a way to get core avc to work with mplayer (in windows and linux) 1080p would be a breeze for virtually all computers because it is so lightweight. It is not so much the codec but the players they are used in.
CoreAVC (at least up to 1.2) is supported in the latest MPlayer SVN revisions. See: http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/2007-February/049188.html
I dont think mplayer is any different than any other media player like zoom player. Unless you can prove me wrong....
MPlayer and VLC are not DirectShow-based players, so they don't play by the same rules. They can't directly use DShow codecs installed on the system, but on the other hand they tend to be a bit faster than other players with their built-in set of codecs. MPlayer (and VLC on Linux) also implement a portion of DirectShow and Win32 API to support binary codecs. However, this support is not perfect, so most new codecs require some tuning and coding before they can be used.
Hans Ohlo
20th March 2007, 22:01
proof: (bsg 1080paff stream)
coreAVC 1.3 (same result with 1.2 but with fixed color bug)
http://img131.imageshack.us/img131/8586/coreavcfuckuppaffed2.png
ffdshow:
http://img131.imageshack.us/img131/5677/ffdshowfinesy0.png
look at the fonts...
i found this surfing some hdtv sites. the caps are from skyoneHD broadcasts of battlestar galactica which are 1080 paff streams. the screenshots clearly show that coreAVC still has big issues correctly decoding some streams. this kind of stream was it that showed the known color levels bug.
so i again ask myself what is happening to coreAVC, is it still the prime product of coreCodec or did they move on to something else?
AKarpo
20th March 2007, 22:17
i found this surfing some hdtv sites. the caps are from skyoneHD broadcasts of battlestar galactica which are 1080 paff streams. the screenshots clearly show that coreAVC still has big issues correctly decoding some streams. this kind of stream was it that showed the known color levels bug.
so i again ask myself what is happening to coreAVC, is it still the prime product of coreCodec or did they move on to something else?
Dude, chill out. They're working on it. If you don't like the product, don't buy it. I've seen you harassing CoreAVC on this thread for like three posts now. We get it.
I'm sure if you were to provide the material to Betaboy so that they may fix it in the next version, they would be more than willing to take a look at it.
ChronoCross
20th March 2007, 22:18
is there any particular point you post in this thread hans ohlo? I mean all you do is whine and moan. If you don't like coreavc then go use ffdshow.
saratoga
21st March 2007, 01:03
i found this surfing some hdtv sites. the caps are from skyoneHD broadcasts of battlestar galactica which are 1080 paff streams. the screenshots clearly show that coreAVC still has big issues correctly decoding some streams. this kind of stream was it that showed the known color levels bug.
so i again ask myself what is happening to coreAVC, is it still the prime product of coreCodec or did they move on to something else?
Are you sure thats not due to different frames being used? Its pretty clear the one image is a good couple frames off from the other. The text may not actually be the same in the source, particularly if it was ripped from a lower bitrate MPEG2 source.
bob0r
21st March 2007, 04:38
Don't use the initial 1.3 release if you want to play:
H.264 display in dvbviewer = fucked (capturing works)
H.264 in .m2ts = fucked
H.264 in .EVO = fucked
Its a good thing they let us "beta test" first, before releasing to the masses :rolleyes:
(Btw, all those options work fine in 1.2, the only thing required for dvbviewer is the crop 1088 to 1080 option, which hides all the mess from the broadcasters)
mikeyakame
21st March 2007, 05:17
1.3 is still producing heavy artifacts/blocking trails on black blocks, where as cyberlink/ffdshow are producing clean outputs, only seems to happen on very high bitrate 1080p encodes, video bitrate avg for this is 10217kbps. would you like a sample?
bob0r
21st March 2007, 05:35
Samples are always good.
BetaBoy
21st March 2007, 06:08
There still is an issue with coreavc creating artifacts with encodes.
http://forum.doom9.org/showthread.php?p=945658#post945658
Im not sure if this will be fixed........
Got your samples.... we are not seeing the artifacts you are seeing... the only to note is that we did see some blockiness on the door.
BetaBoy
21st March 2007, 06:09
Samples are always good.
I agree... samples please. Thx
BetaBoy
21st March 2007, 06:53
'less CPU load more ppl can play 1080p h264 or more sofisticated decoder can be build '
If there was a way to get core avc to work with mplayer (in windows and linux) 1080p would be a breeze for virtually all computers because it is so lightweight. It is not so much the codec but the players they are used in.
Hey betaboy, if you how to make this possible, or know of a player as light as mplayer, let me know. Hopefully someone has cracked it.
By the way i have tried TCMP, kmplayer,bsplayer etc .
Both CorePlayer Pro and BetaPlayer for Linux will be out by the end of the year.
mrkazador
21st March 2007, 07:05
Got your samples.... we are not seeing the artifacts you are seeing... the only to note is that we did see some blockiness on the door.
Wow...Really?
I dont see how it not could be coreavc, FFDSHOW works prefectly fine. Im using Haali Media Splitter, also tried Gabest matroska splitter and same problem.
Im not saying your a liar :devil: but can anyone else confirm that it is fixed? Could be on my end.
More info on the problem here;
http://forum.doom9.org/showthread.php?t=119187&highlight=coreavc+bug
Sample here;
http://rs38.rapidshare.com/files/21985628/Doom9.rar
Hans Ohlo
21st March 2007, 07:50
I mean all you do is whine and moan. If you don't like coreavc then go use ffdshow.
i did not pay for ffdshow. i payed for coreAVC and get only lame excuses and they tell me that everything will be fixed in a release that gets delayed and delayed. then it finally shows up and nothing is fixed (as many here point out).
I'm sure if you were to provide the material to Betaboy so that they may fix it in the next version, they would be more than willing to take a look at it.
if he would react to a pm once i would try to get the sample for the pictures. and i am on it.
Are you sure thats not due to different frames being used? Its pretty clear the one image is a good couple frames off from the other. The text may not actually be the same in the source, particularly if it was ripped from a lower bitrate MPEG2 source.
nope, this is the h264 paff source stream. as i said i will try to get a sample.
BUT how long will it be till this is finally fixed? another 6 or so months?
and yes lately on thins thread i had nothing good to say but thats only because there simply are only bad things around coreAVC lately.
ChronoCross
21st March 2007, 15:32
i did not pay for ffdshow. i payed for coreAVC and get only lame excuses and they tell me that everything will be fixed in a release that gets delayed and delayed. then it finally shows up and nothing is fixed (as many here point out).
and yes lately on thins thread i had nothing good to say but thats only because there simply are only bad things around coreAVC lately.
you also haven't provided anything to help fix the problem. You've done a whole lot of complaining that is best meant for the ranting thread. Pictures of the problem don't count as how do you fix a picture.
Hans Ohlo
21st March 2007, 16:52
ok then:
http://www.megaupload.com/?d=7S6FA5K3
lets see how long this will take ;)
mrkazador
21st March 2007, 20:11
ok then:
http://www.megaupload.com/?d=7S6FA5K3
lets see how long this will take ;)
I captured the same frame you posted before. The only problem I see is with "skip all deblocking" you get some kind of blocking/artifacts near the text. With standard deblocking this goes away.
Coreavc with standard deblocking:
http://img103.imageshack.us/img103/104/coreavcstandarddeblockigy4.th.jpg (http://img103.imageshack.us/my.php?image=coreavcstandarddeblockigy4.jpg)
Coreavc with Skip ALL deblocking:
http://img240.imageshack.us/img240/2063/coreavcnostandarddeblocea1.th.jpg (http://img240.imageshack.us/my.php?image=coreavcnostandarddeblocea1.jpg)
Using ffdshow, it produced the exact same image as "standard deblocking"
Hans Ohlo
21st March 2007, 21:01
I captured the same frame you posted before. The only problem I see is with "skip all deblocking" you get some kind of blocking/artifacts near the text. With standard deblocking this goes away.
Coreavc with standard deblocking:
http://img103.imageshack.us/img103/104/coreavcstandarddeblockigy4.th.jpg (http://img103.imageshack.us/my.php?image=coreavcstandarddeblockigy4.jpg)
Coreavc with Skip ALL deblocking:
http://img240.imageshack.us/img240/2063/coreavcnostandarddeblocea1.th.jpg (http://img240.imageshack.us/my.php?image=coreavcnostandarddeblocea1.jpg)
Using ffdshow, it produced the exact same image as "standard deblocking"
hi, thanks for checking.
ok i found out the following:
- no deinterlacing, the picture is fine***
- software deinterlacing, the picture looks as in my screenshots
***the colorbug seems to be back, convert TV to PC scale OFF (i have a calibrated output), i checked this with ffdshow behind coreAVC in the filtergraph. the levels are off (not 16-235, more like 32-119). just scaling to PC levels would result in correct TV scales. so color precision is lost completely.
please could someone check my findings.
mrkazador
21st March 2007, 21:22
***the colorbug seems to be back, convert TV to PC scale OFF (i have a calibrated output), i checked this with ffdshow behind coreAVC in the filtergraph. the levels are off (not 16-235, more like 32-119). just scaling to PC levels would result in correct TV scales. so color precision is lost completely.
please could someone check my findings.
Your right about this, I had to "check" tv to pc scale. I think betaboy knows about this problem already.
Hans Ohlo
21st March 2007, 22:29
Your right about this, I had to "check" tv to pc scale. I think betaboy knows about this problem already.
could you also check if when you use the point software deinterlacing the picture looks as in the first screenshots i quoted?
DeepBeepMeep
21st March 2007, 23:51
I have reported the colorspace issue in many occasions (see my two video samples) and didn't get any feedback so far )-:
I don't need to turn on deinterlacing to get the wrong color space with black level at 30. The black level has been fixed for some hd video sources (some stuff coming from PremiereHD have black level at 16 without using the new option).
However, hd video material coming from SkyHD have both incorrect black levels and white levels (black at 30 et white at 226 instead of being respectively at 16 and 240 for TV levels).
The new option at best enlarges slightly the color space but doesn't do what do what its name says. And as I mentioned previously since the issue has been fixed only partially the option needs sometime to be turned on sometime turned off.
Now betaboy has all he needs to fix the issue. It would be good at least to get an official feedback from him to know if his is aware of the issue and working on it...
mrkazador
22nd March 2007, 00:43
could you also check if when you use the point software deinterlacing the picture looks as in the first screenshots i quoted?
Yep, same thing. Software deinterlacing adds some sort of artifacts in the text, like lines.
Hans Ohlo
22nd March 2007, 07:26
However, hd video material coming from SkyHD have both incorrect black levels and white levels (black at 30 et white at 226 instead of being respectively at 16 and 240 for TV levels).
The new option at best enlarges slightly the color space but doesn't do what do what its name says. And as I mentioned previously since the issue has been fixed only partially the option needs sometime to be turned on sometime turned off.
actually the source material does not have incorrect levels, coreAVC outputs incorrect levels at these kind of streams.
the TV option is no option for me, because with it color precision is lost. it only stretches the levels...
another thing i noticed... corecAVC ;) and many typos in the properties tab. lets you wonder how high the priority of this project is at corecodec.
Yep, same thing. Software deinterlacing adds some sort of artifacts in the text, like lines.
thanks for confirming... now lets wait a year and see if it will be fixed.
don't understand me wrong, i like coreAVC. it is the only decoder which plays some streams of mine fluently but lately the issues come up and get totally ignored by the devs. i/we payed for the software!
DeepBeepMeep
22nd March 2007, 09:37
actually the source material does not have incorrect levels, coreAVC outputs incorrect levels at these kind of streams.
Sorry I wasn' t very clear. I meant exactly what you said: colors levels rendered for these streams are incorrect with coreavc whereas they are fine with ffdshow, cyberlink h264 decoder, etc...
Seb.26
22nd March 2007, 12:45
About the "TV levels bug" ...
If you take a TV level source ( a DVD ie. ) ... them encode it with x264 or same, the levels "in the encoded file" are identical ( encoding doesn't re-map levels ?!! ) ... and them decode this with CoreAvc, levels aren't the same ?!
... In another words : does CoreAvc do a PC->TV level ?! ...
acidsex
22nd March 2007, 15:38
Has 1.3 been released yet? I havent received my info and no new version is listed under my core account. Is there a cost to upgrade to 1.3?
Selur
22nd March 2007, 18:15
just download the file in your account again. I did this and when installing the new file it states 1.3 ;)
acidsex
22nd March 2007, 18:38
Cool, got the file, went to install and it told me it has already been activated. i thought we were supposed to be able to install this on Vista now? Do I have to deactivate my previous license used under XP? Surely Im not expected to purchase a new license for use with Vista when the old license is no longer being used. Any suggestions?
Haali
22nd March 2007, 23:55
I have reported the colorspace issue in many occasions (see my two video samples) and didn't get any feedback so far )-:
However, hd video material coming from SkyHD have both incorrect black levels and white levels (black at 30 et white at 226 instead of being respectively at 16 and 240 for TV levels).
I've analyzed your two samples, and they both have full range specified in SPS VUI, but are using only 16-235 levels. I think CoreAVC behaves correctly and respects VUI information that says the video is already full range. 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.
The first sample is also complicated because it is a Matroska file renamed to ts, and has VIU info stripped from the SPS stored in Matroska headers, but the stream has more SPS NALs with VUI in place that specify full range. For some time those extra SPS were removed by the splitter, but this won't be done anymore because I now have samples where SPS/PPS does change in the middle of a stream and has to be kept.
DeepBeepMeep
23rd March 2007, 01:25
Haali, thanks for your explanation.
I have indeed mistakenly sent a sample which indeed came from a mkv file. The ironic thing is that once turned into a mkv file the color levels are good (maybe as you suggested because some info was lost).
The fact that it is a mkv file explains why the colors looked good on one sample and not on the other one. In fact I checked a TS sample from the same broadcaster of the sample that was turned into a mk file and when it is still in its TS form it has also incorrect color levels.
So it seems the "full range" flag is improperly set for two of the major european HD broacasters (SkyHD and PremiereHD). Since I assume most HD content created are stored no matter what using TV levels, it would be very nice if CoreAVC could ignore the flawed flag like it is done by other decoders such as ffdshow and cyberlink.
This could be an option if required. It seems this would fix the issue reported by Han Solo and others. Many thanks from the CoreAVC team if you could do something about it because at the moment it is a bit annoying to have to turn on and off an option to watch a movie.
vBulletin® v3.8.11, Copyright ©2000-2025, vBulletin Solutions Inc.