Media Player Classic supports Matroska v2, *.mp4, *.mov, and MPEG-4 Timed Text

[Yong]

As usual, thx Celtic druid, for the latest version of mpc :) and Liisachan, for the heads up :D
btw, looks like celtic druid forgot to include the manifest file(?) when compiling mpc, i have to manually put a mplayerc.exe.manifest file in same dir in oder to restore the "Windows XP looks and feel" :p

[rotflol]

Is it just me or does image rotation not work in the latest build?

[Egh]

Technically tho, VSFilter's internal algo is sub-pixel too when, for instance, antialiasing. That's why you can specify non-integer border width in ASS.


So, basically, when softsubbing, we should try not to be too dependent on a specific sub renderer's default settings... It's just like we shouldn't make a web page that works only for IE and only when font size = small. I guess quite a few people use mplayer on Linux, or other sub renderers, so it'd be pointless even if VSFilter and MPC were 100% compatible. In other words, if absolute positioning is critical, you might want to hardsub that part.


First, i'm quite aware of all those techical details, what i really can't find is some info on sub-pixel precision in MPC and VSFilter. How much supersampling is done? Is it 2x or ... ? Is the implementation different in VSFilter and MPC renderers?

Second, changing styles in mpc dialog is annoying. First, it's hard to change many styles at once (and there could be easily 5-10 of them!). Second, inspite of my early bug reports, the problem with shadow and outline width being allowed to have only interger values in mpc styles dialog remain! :mad:

btw -- the mighty haali splitter can scale all the fonts in splitter with % precision.

Third -- hardsubs are bad ;) So hardsubbing, especially non-karaoke parts is pretty much out of question. I'm delighted that subs can be changed on-the-fly (but see the bug above!), but taking your example, i would like to have canonical page which is displayed under same standard settings in an identical way. Which is what actually webdesigners are obliged to achieve (but not always do so :P) It seems that in the end you'll be forced to provide two versions of softsubs for two different renderers for them to be entirely same. I like MPC renderer, btw, it's my personal choice. But many don't use it for various reasons and use VSFilter, which actually comes from same developer :)

@Yong: wow, so you managed to achieve XP look? Cause that problem persists for half a year for builds from celtic. I actually tried in september or so to do same (put manifest in mpc folder) but eventually failed :P Can you upload your manifest file somewhere? Would like to try it :)

[SeeMoreDigital]

Firstly...

Thanks to all those involved in getting MPC to display the correct AAC channel information ;)

http://img50.imageshack.us/img50/5872/aacinmpcb3zh.png


However, I have noticed that when playing AAC ADTS .AAC streams, the files "bit-rate" is not being displayed correctly: -

http://img50.imageshack.us/img50/2752/aacinmpca5fo.png


Cheers

[Yong]

@Yong: wow, so you managed to achieve XP look? Cause that problem persists for half a year for builds from celtic. I actually tried in september or so to do same (put manifest in mpc folder) but eventually failed :P Can you upload your manifest file somewhere? Would like to try it :)

Here you go: ;)
http://www.geocities.com/y0ngc/mplayerc.exe.zip
Actually im copy the manifest files from others apps, then rename it to mplayerc.exe.manifest, im feel suprise its worked :p

[Liisachan]

How much supersampling is done? Is it 2x or ... ? Is the implementation different in VSFilter and MPC renderers? I can't really code dive, but VSFilter was originally Vobsub, which was based on subtitler.vdf's Rasterizer, whose readme says: "Subtitler antialiases all fonts by default at an 8x8 setting, i.e. it renders
*all* glyphs at eight times normal size and reduces it down."

hardsubs are bad ;) So hardsubbing, especially non-karaoke parts is pretty much out of question. Don't tell it to me ;) I am one of the world first real multisubbers. But like you said, the best way is probably "mixsub" i.e. softsub+little hardsub for special effects.

[SeeMoreDigital]

I have a feature request... if I may be so bold ;)

Given that MediaPlayer Classic already offers on-board AAC, AC3 and DTS decoders, would it be possible to include 6Ch AAC to 6Ch AC3 transcoding. Or even 6Ch AAC to 6Ch DTS transcoding?

Just a thought ;)

[Egh]

I can't really code dive, but VSFilter was originally Vobsub, which was based on subtitler.vdf's Rasterizer, whose readme says: "Subtitler antialiases all fonts by default at an 8x8 setting, i.e. it renders
*all* glyphs at eight times normal size and reduces it down."

Don't tell it to me ;) I am one of the world first real multisubbers. But like you said, the best way is probably "mixsub" i.e. softsub+little hardsub for special effects.

8x8 supersampling is a bit redundant imo, 4x4 would be more than just fine. So are you sure it's really 8 times in *each* direction? Cause it's 64 pixels downsampled to one, a bit too much I'd say.

And as for identical/nonidentical results, i wasn't talking about just special effects like karaoke. I was talking about general typesetting tricks for main episode.

[Liisachan]

8x8 for hq antialiasing is normal isnt it? Or rather, it might be float, not x8 int.
You can test it by yourself...if qpel, 3 or more settings between border=1.0px and 1.25px would be quantized to 2 results. If border=1px 1.1px 1.2px 1.3px are all different, then algo is somehow more subpixel than qpel.

Another interesting testing would be {\p2} or 2+
Gabest himself writes:
"For example {\p4} and the coordinate (8,16) will mean the same as {\p1} and (1,2). This feature can be useful for sub-pixel accuracy."

[Egh]

Actually im copy the manifest files from others apps, then rename it to mplayerc.exe.manifest, im feel suprise its worked :p

Respect o.O It actually works :approved: So finally I have winXP controls back. Just wondering why it didn't work when I tried same thing as you...

[B.F.]

I can't really code dive, but VSFilter was originally Vobsub, which was based on subtitler.vdf's Rasterizer, whose readme says: "Subtitler antialiases all fonts by default at an 8x8 setting, i.e. it renders
*all* glyphs at eight times normal size and reduces it down."

8x8!!!
Now I know why CPU load is so high.
How about a manual control?

[Egh]

8x8!!!
Now I know why CPU load is so high.
How about a manual control?

Yeah, could be nice option. In fact, it's not a big problem when you have prebuffering on (3 pix ahead is my typical choice). But when you have karaoke, moving signs and other stuff like that you have to turn buffering off and then CPU % for subs sky-rockets :)

So essentially I would decrease supersampling (probably to 2x2 only) for karaoke and moving signs.

[Stebs]

(About Haali renderer) Last modified Tuesday, 21 February 2006 04:19:06 AM
Still says v0.1I suppose it is still non-public, or somebody has a link? Cant wait to test this ;)

The open file dialogue has a dub option where you can load audio.DOH! :o Have I been THAT blind all the time? -Well anyway, many thanks! :D

[Elias]

I tried Windows Vista Beta Build 5308 yesterday, and MPC worked great :) However, Vista was still buggy :D Anyone else tried MPC on Vista? I could select all file formats in options, but AIFF Format Sound couldn't be checked. But I guess that's Vista's fault.

[Egh]

I suppose it is still non-public, or somebody has a link? Cant wait to test this ;)



Sneak preview :O DXR is included in next release of Haali splitter, so installing that you'll get his renderer as well :p

The next release is close to the end of testing phase, so soon will be out.

[Stebs]

Great :), thanks for the info!

[Liisachan]

mplayerc2006.02.26.2kxp.7z (http://www.fairyland.to/celtic_druid/Media_Player_Classic/mirror/Media%20Player%20Classic/mplayerc2006.02.26.2kxp.7z)
updated: MP4Splitter.7z (http://ffdshow.faireal.net/mirror/Media%20Player%20Classic/external%20filters/unicode/MP4Splitter.7z)

[sillKotscha]

QT bug (SQV3+mp3 in mov) is still prensent in the latest build ...

not anymore :)

many thanks for your efforts ... very nice !!

unfortunately the mov container still remains to be evil...

the good old 'tyson bite' mov-clip:

audio: mono adpcm_ima_qt @ really low 11025 Hz
video: Cinepak (CVID)

maybe it is a ffdshow_audio decoding bug which the mp4 splitter can't handle very well

- if you are interested, I have uploaded (http://s65.yousendit.com/d.aspx?id=3H21U38D0LUIT2Q6A4RXBPE3PN) the clip (718 KB).

thank you

EDIT:

I have another mov clip with the same attributes except audio (8000 Hz instead of 11025 Hz) that file plays just fine...

[Tima]

Thanks for fixing svq3+mp3! =)

On any mov file I get this warning:

E:\Video\!Trailers\xxx2_superbowl_1024.mov::Apple Video Media Handler

Media Type 0:
--------------------------
AM_MEDIA_TYPE: (Video: SVQ3 1024x576)
majortype: MEDIATYPE_Video {73646976-0000-0010-8000-00AA00389B71}
subtype: Unknown GUID Name {33515653-0000-0010-8000-00AA00389B71}
formattype: FORMAT_VideoInfo {05589F80-C356-11CE-BF01-00AA0055595A}
bFixedSizeSamples: 1
bTemporalCompression: 0
lSampleSize: 1
cbFormat: 211

VIDEOINFOHEADER:
rcSource: (0,0)-(0,0)
rcTarget: (0,0)-(0,0)
dwBitRate: 0
dwBitErrorRate: 0
AvgTimePerFrame: 0

BITMAPINFOHEADER:
biSize: 40
biWidth: 1024
biHeight: 576
biPlanes: 0
biBitCount: 0
biCompression: SVQ3
biSizeImage: 0
biXPelsPerMeter: 0
biYPelsPerMeter: 0
biYPelsPerMeter: 0
biClrUsed: 0
biClrImportant: 0

pbFormat:
00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000030: 28 00 00 00 00 04 00 00 40 02 00 00 00 00 00 00
00000040: 53 56 51 33 00 00 00 00 00 00 00 00 00 00 00 00
00000050: 00 00 00 00 00 00 00 00 00 00 00 7b 53 56 51 33
00000060: 00 00 00 00 00 00 00 01 00 03 00 00 53 4d 49 20
00000070: 00 00 00 00 00 00 02 00 04 00 02 40 00 48 00 00
00000080: 00 48 00 00 00 00 00 00 00 01 10 53 6f 72 65 6e
00000090: 73 6f 6e 20 56 69 64 65 6f 20 33 00 00 00 00 00
000000a0: 00 00 00 00 00 00 00 00 00 00 00 20 ff ff 00 00
000000b0: 00 0c 67 61 6d 61 00 02 80 00 00 00 00 15 53 4d
000000c0: 49 20 53 45 51 48 00 00 00 05 e8 00 48 19 c0 00
000000d0: 00 00 00


Media Type 1:
--------------------------
AM_MEDIA_TYPE: (Video: SVQ3 1024x576)
majortype: MEDIATYPE_Video {73646976-0000-0010-8000-00AA00389B71}
subtype: Unknown GUID Name {33717673-0000-0010-8000-00AA00389B71}
formattype: FORMAT_VideoInfo {05589F80-C356-11CE-BF01-00AA0055595A}
bFixedSizeSamples: 1
bTemporalCompression: 0
lSampleSize: 1
cbFormat: 211

VIDEOINFOHEADER:
rcSource: (0,0)-(0,0)
rcTarget: (0,0)-(0,0)
dwBitRate: 0
dwBitErrorRate: 0
AvgTimePerFrame: 0

BITMAPINFOHEADER:
biSize: 40
biWidth: 1024
biHeight: 576
biPlanes: 0
biBitCount: 0
biCompression: svq3
biSizeImage: 0
biXPelsPerMeter: 0
biYPelsPerMeter: 0
biYPelsPerMeter: 0
biClrUsed: 0
biClrImportant: 0

pbFormat:
00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000030: 28 00 00 00 00 04 00 00 40 02 00 00 00 00 00 00
00000040: 73 76 71 33 00 00 00 00 00 00 00 00 00 00 00 00
00000050: 00 00 00 00 00 00 00 00 00 00 00 7b 53 56 51 33
00000060: 00 00 00 00 00 00 00 01 00 03 00 00 53 4d 49 20
00000070: 00 00 00 00 00 00 02 00 04 00 02 40 00 48 00 00
00000080: 00 48 00 00 00 00 00 00 00 01 10 53 6f 72 65 6e
00000090: 73 6f 6e 20 56 69 64 65 6f 20 33 00 00 00 00 00
000000a0: 00 00 00 00 00 00 00 00 00 00 00 20 ff ff 00 00
000000b0: 00 0c 67 61 6d 61 00 02 80 00 00 00 00 15 53 4d
000000c0: 49 20 53 45 51 48 00 00 00 05 e8 00 48 19 c0 00
000000d0: 00 00 00




When I close the warning, everything plays fine, with video and sound. :)

QT Alternative is not installed.

[sillKotscha]

re-install CoreAVC and it might be fixed...

[Tima]

I use FFDShow, not CoreAVC.

BTW, could you give me a link to dl CoreAVC? ;)

[Liisachan]

Misc (not by celtic_druid)/ (http://ffdshow.faireal.net/mirror/)

[rotflol]

Bug report(?): subtitles like

{1249}{1249}Something something or other etc.

don't work in MPC. They do seem badly formed (the same beginning and ending frame), but they work in ffdshow and VLC.

[Liisachan]

you mean something like
Dialogue: 0,0:00:05.00,0:00:10.00,style1,Alice,0000,0000,0000,,{1249}Hi!
?
{ and } are special characters for SSA/ASS, like < and > in HTML,
and last time I checked they are special even in SRT for Gabest's parser.

The workaround, if it's ok for you to write the script in Unicode, would be to use U+FF5B ｛ and U+FF5D ｝ for ASCII 0x7B { and 0x7D } respectively.

[Liisachan]

mplayerc2006.02.26-2.2kxp.7z (http://www.fairyland.to/celtic_druid/Media_Player_Classic/mirror/Media%20Player%20Classic/mplayerc2006.02.26-2.2kxp.7z)

Maybe OggSplitter is updated?

EDIT: Hm but Theora.ogg still doesn't play properly by MPC itself.
And if OggSplitter is disabled, + Haali = works, but + illiminable = crash.

[Reino]

MOV playback results of MPC (18022006) + FFDShow:

MOV[AVC1+AAC] & MOV[MP4V+AAC] --> good.
MOV[SVQ3+MP3] --> CPU: normal, audio: a little better but still bad and still glitches, video: a little better but still bad; very choppy.
MOV[SVQ3+AAC] --> the same.
MOV[SVQ3+TWOS] --> the same.
MOV[SVQ3+QDM2 or IMA4] --> the same.
MOV[SVQ1+IMA4 or ARAW] --> the same.
MOV[CVID+TWOS or IMA4] --> CPU normal, audio: ... (available?), video: good, time seek: good.

MOV playback results with the Gabest MP4 Splitter (25022006) + FFDShow:

MOV[SVQ3+MP3 or AAC] --> audio-, video playback, and timeseek: very good!
(allthough one little thing: noitulove.mov hangs at the end when it's about to playover again (looping); no problems with the Crowd Demo though)
MOV[SVQ3+TWOS] --> with MPC the internal MP4 Splitter is used of course and so only video is rendered.
with Zoom Player the Quick Time Movie Parser (quartz.dll) kicks in, and it seems TWOS is just PCM audio. It plays fine!
MOV[CVID+TWOS] --> unlike above, where the Quick Time Movie Parser kicked in, this doesn't happen here, so only video playback again. (perhaps this filter isn't capable of splitting CVID and thus the MP4 Splitter does the job).
MOV[SVQ1 or SVQ3+IMA4] --> audio- (stereo only!), video playback, and timeseek: very good!
(note: mono IMA4 sounds terrible though :p)
MOV[SVQ3+QDM2] --> audio-, video playback, and timeseek: very good!
MOV[SVQ1+ARAW] --> good, but just no audio support from FFDShow yet :(


MOV files I used:
MOV[SVQ3+MP3] -->
-http://pcmovies.ign.com/pc/video/doom3d/doom3_051004_qthigh.mov (73,3MB)
-http://www.stormahead.com/images/STORMAHEAD_PodsTrailer.mov (7,72MB)
MOV[SVQ3+AAC] -->
-http://media.cgnetworks.com/cgfilms/ads/noitulove.mov (6,76MB)
-http://www2.ati.com/misc/demos/ATI-X800-Crowd-Demo-v1.0.mov (55,8MB)
MOV[SVQ3+TWOS] -->
-http://downloads.gamezone.com/demos/d12219.htm or http://www.3ddownloads.com/Action/DOOM%203/Movies/Resurrection%20of%20Evil%20trailer.mov;jsessionid=413D406550B1C10C0F5E04842373399E (40,0MB)
MOV[SVQ3+QDM2] -->
-http://www.slothy.com/casting.mov (1,28MB)
MOV[SVQ1+ARAW] -->
-http://www.hkpro.com/video/MP5.mov (3,66MB)
MOV[CVID+TWOS] -->
-http://rapidshare.de/files/12877939/dolby_city__CVID_TWOS_.mov.html (5,55MB)
MOV[SVQ1+IMA4] -->
-sillKotscha's posted MOV (http://forum.doom9.org/showthread.php?p=790872#post790872) (mono)
-http://absinth.dyndns.org/wolf/goodies/MO_130.mov (129MB, stereo)

Very impressive update on the MP4 Splitter! :thanks:
Too bad the OGG Splitter still has its issues with THEORA and Vorbis timeseeking.
And also the MPEG Splitter has some difficulties with certain MPEG1 Video files concerning timeseeking. (the video hangs, sound goes on)

[SeeMoreDigital]

Just tested the following mplayerc2006.02.26-2.2kxp.7z (http://www.fairyland.to/celtic_druid/Media_Player_Classic/mirror/Media%20Player%20Classic/mplayerc2006.02.26-2.2kxp.7z) build with several video only SVQ3 in MOV files.... They all appear to play perfectly, totally free from stutter now ;)

I never thought I'd be dedicating so much of my time checking to see if video (and audio) streams within "Apple's MOV container" function correctly....

It is however quite good fun :)


Cheers

[Peuj]

Hi,

With the latest build, the auto-load subtitle doesn't work (at least for me). I've the problem with a simple srt file (same name as the video file)

I've tried with latest official version and it works.

Can someone confirm it's a bug ?

Thanks

[rotflol]

you mean something like
Dialogue: 0,0:00:05.00,0:00:10.00,style1,Alice,0000,0000,0000,,{1249}Hi!
?
{ and } are special characters for SSA/ASS, like < and > in HTML,
and last time I checked they are special even in SRT for Gabest's parser.
No, they're .txt subs in mdvd format. I should've made myself clear.
{number1}{number2}Text

Numbers 1 and 2 tell when the line is supposed to appear and disappear, respectively. In this case both numbers are the same. The subs were probably converted from some other format which only gives the starting times for subs.

[Avish]

New matroska splitter is out. It contains Haali's Video Renderer. I tried it out with MPC but I'm not able to see any picture...ony sound. Why is that? Is there any special requirements for Haali's Video Renderer to work?

[B.F.]

On my Rageon9550 all work fine, but that render is 2x slower compare to MPC WMR9. :( Maybe on my home PC with GF6600 it will be better.

In new MPC version NumPad bittons doesn't work correctly.

[Liisachan]

Now "Haali's Video Renderer" is no more grayed out and works for me for many formats except .mp4. I don't see abnormal CPU load, but opening file sometimes takes a while. Once it's opened, it goes normally.

@rotflol sorry I know little about anything except ssa/ass.

[zlab]

Hi,
I'm using media player classic (20060226), when playing a MKV file, it does not respect the display width/height setting, ex: 720*480 suppose to be display as 848*480, VLC do it properly, my friend's mpc also can do that with the same MKV file, could anyone show me the tricks on setting step by step to do that? I've been fool around for awhile but can't figure out how to do that.

[Elias]

Hi,
I'm using media player classic (20060226), when playing a MKV file, it does not respect the display width/height setting, ex: 720*480 suppose to be display as 848*480, VLC do it properly, my friend's mpc also can do that with the same MKV file, could anyone show me the tricks on setting step by step to do that? I've been fool around for awhile but can't figure out how to do that.This is a decoder issue. You have to enable this in ffdshow. This question has been answered in this thread before.

[Avish]

Now "Haali's Video Renderer" is no more grayed out and works for me for many formats except .mp4. I don't see abnormal CPU load, but opening file sometimes takes a while. Once it's opened, it goes normally.
Can anyone tell me why "Haali's Video Renderer" doesn't work with MPC on my comp? I tried every video format but I can only hear sound with a black screen. Im using the latest versions of both. What am I missing?

[kurt]

on my pc it doesn't work too (geforce 4 Ti 4200), maybe we need another X-card? :)

However, for those who don't have at least 6xxx series G-Force or equivalent, new renderer might be not much of use due to low speed.
klick (http://forum.doom9.org/showthread.php?p=776380#post776380)

[zlab]

This is a decoder issue. You have to enable this in ffdshow. This question has been answered in this thread before.

Hi, Thanks for yr info. My findings seen interesting.
It's probably due to ffdshow setting.
When the raw video set to "all supported" - AR incorrect.
When the Overlay - Output, choice "Use Overlay mixer" then whatever in the raw video the output is now AR corrected.
It is using CoreAVC as decoder, ffdshow for postprocessing. AR now corrected.

[B.F.]

Tested Haali Render on home PC.
Cool.
Video+Subtitles work mutch faster now. :)

[clsid]

Perhaps the Haali renderer needs a graphics card with PixelShader 2.0?

[Avish]

Info on Haali renderer: seems it's a subvariant of VMR9, and completely GPU based, i.e. operates only if shaders supported by 3D card. It's believed that the renderer now incorporates much-expected faster and proper version of bicubic resizer (compared to old MPC one, which is written by Haali as well :P).
@clsid It seems you are right. I guess I'll have to buy a new card :rolleyes:

[Isochroma]

Works here, but I get 99-100% CPU usage. Using an FX5200, XP 2500+

[Stebs]

With my 6800GT, Haali Renderer is still a bit slower than VMR9. Also, as others already reported, opening a video file needs some seconds until something happens.
Got some crashes with HD wmv files (like Amazing_Caves_WMV_HD_1080p.wmv).

[Liisachan]

PS 2.0 Resizing problem (http://forum.doom9.org/showthread.php?p=745819#post745819) (VMR9 Renderless + Bilinear (PS 2.0) = Abnormal CPU load) is fixed, if I use "Haali's Video Renderer."

My card is not very new, but supporting PS 2.0a. It's Quadro FX 500 and Stebs told me "it's based on the NV34GL Chip which is similar to a GeforceFX 5200" in case that info helps.

[Egh]

Works here, but I get 99-100% CPU usage. Using an FX5200, XP 2500+

@Avish and all concerning this issue:

in fact for big frames i can't get it run fast enough even on +10% OCed G5500FX. And even on smaller frames it's slower than pure software mode in VMR9.

And that's not comparing with just frame, i mean comparing with original frame decoding + *software* bicubic/lanczose resizing full screen (e.g. 720*480 to 1024*576) and then rendering it with VMR9 (but in mixer mode). IIRC in mixer mode he output can be YV12 btw! So no CPU is used for converting colorspaces.

So my original point (based on very early alpha-test build) so far stands: haali renderer is not good for anything less than 6xxx GF.

[breez]

Haali renderer:

- CPU usage a tad more stable than VMR9 when not resizing (avg the same)
- When resizing, noticeable higher CPU usage (compared to VMR9 + PS 2.0 bicubic)
- Does quality colorspace conversion to RGB32 with levels expanded to 0-255 from 16-235, but CPU usage is a tad higher and less stable than VMR9 + ffdshow HQ RGB32 conversion

* On a Radeon 9800Pro (PS 2.0)

[Egh]

Haali renderer:

- CPU usage a tad more stable than VMR9 when not resizing (avg the same)
- When resizing, noticeable higher CPU usage (compared to VMR9 + PS 2.0 bicubic)


Original PS 2.0 is slow as hell because it's very simpistic and does it upsampling in a straightforward way. Haali renderer is more sophistcated and is supposed to be faster with everything else same.

[breez]

Original PS 2.0 is slow as hell because it's very simpistic and does it upsampling in a straightforward way. Haali renderer is more sophistcated and is supposed to be faster with everything else same.

On my system the PS 2.0 bicubic resize provided by MPC has zero CPU hit. With Haali's renderer there is extra 10-20% CPU used (1900MHz Athlon XP). Of course the renderer is still under development and there might be some problems with it. I will try the renderer with different MPC versions.

edit: With the latest MPC found in this thread the CPU usage is _very_ stable with Haali renderer when resizing, on average only around 5-6% more than with VMR9 + PS 2.0 bicubic resize.

edit2: Results vary between different video files.

The above results I got with a 640x480 XVID in AVI. Similar results with a bit higher CPU hit with a 768x432 XVID in MKV, but a 20%+ CPU hit with a 720x384 XVID in MKV. Also, 100% CPU usage with a 640x480 WMV9 in AVI when resizing.

[Rash]

I've also noticed Haali has some tearing, depending on the movement (and the framerate), though I believe the general image quality is better.

[Dreamer-FLT]

Since cvs 02/16/2006 broken is the playback of an *.ac3 audio, if no internal splitter is used ("DTS/AC3" is *disabled*). It seems that newer cvs shots aren't able anymore to connect to the standard ac3 parser filter (mpg2splt.ax).

This is not dependent of any buggy external splitter/demuxer. I tried almost all of them, and the official MPC builds are working. As an side effect MPC hangs too, if "auto load audio files" is optioned and an external *.ac3 file exists.

Its clear that cvs shots aren't for public usage, but i think it has to be fixed for the final release.

Thanks for listening,
Dreamer

[Liisachan]

bicubic, with Haali. is now working for me too, with around 20~30% CPU load. 30% is a bit expensive, but considering that Bicubic+VMR 9 Renderless is not "affordable" at all for my box (I get CPU 100%, old P4 3.4G), that's awesome :D At least technically... SO this is what this was all about?

Hi Liisachan,

a good news, this is the answer from Gabest about our problem:
"Yes, it's really bad, but I made the fixes already. There's also going to be a lot faster two pass bicubic resizer in the next version.


:)