ffdshow tryouts project: Discussion & Development

[chros]

Originally Posted by drevil_xxl

It seems that is broken.No multithreaded.

Woowww!!! At least some progress in this topic...
I'm hardly waiting a working version to test ...

[Inventive Software]

Originally Posted by drevil_xxl

Tested with:
http://www.mplayerhq.hu/MPlayer/samp...s/AVC1/bb2.mp4
1 core:
User: 17s, kernel: 0s, total: 17s, real: 17s, fps: 202.5, dfps: 202.5

2 cores:
User: 0s, kernel: 0s, total: 0s, real: 17s, fps: 4365.5, dfps: 208.6

It seems that is broken.No multithreaded.

4365 FPS is mighty impressive, even if it's broken completely!

[Jeremy Duncan]

Using xxl 1240 FFDshow
http://downloads.sourceforge.net/ffd...5&big_mirror=0

Here's the log
http://files.filefront.com//;7690920;;/

Here's a picture of the graph


I used FFDshow Mpeg2 libavcodec dvd decoding and the nvidia video decoder.

It still showed no back or ahead, buffer was checked in Avisynth.


i USED vmr9, DEFAULT OUTPUT IN mpc.
The plugins I linked to before.
and this code

#debug
SetMemoryMax(1024)
FluxSmoothT(3)
ColorMatrix(Mode="Rec.601->Rec.709",scaling=2,threads=2)
MT("SeeSaw(NRlimit=0, NRlimit2=5, Sstr=2.40, Szp=1, Slimit=50, Sdamplo=16, Spower=1, sootheT=95, SdampHi=39)",3)
lanczosresize(720, 480)

Edit. I'm using Vista Premium.

Hi,

Sorry if this is the wrong place to post this question about ffdshow usage, but I'm having trouble finding the right place with so many different pages out there. If it is, can anyone point me to the correct place?

I've used a lot of different recent builds by various people with different results, and am wondering what is the best build for my particular setup, as I'm aware they are sometimes optimized for various Operating systems and instructions sets.

I have an old PIII 1.4ghz with MMX and SSE but nothing above that, and am running XP SP2.

I am currently using the ffdshow_rev1099_20070414_xxl.exe build.

Any advice on the best build to use, and where to get it would be appreciated.

Matt

[clsid]

Any of the generic builds will do just fine.

[chros]

I have been playing around with the new Leak's patched ffdshow (ffdshow_rev1237_20070603_clsid.exe), and try to accomplish this:

Is it makes sense at all to telecine 23.976 fps progressive content to 59.94 fps progressive one on a LCD TV/monitor running 60 Hz (from my comp) ?

If it is:
I have tried a 720p-avc content, with MPC, ffdshow decodes H264 and Haali's renderer. Settings:
x Avisynth , Ignore Pulldown, Buffer back/ahead: 30/30 (I don't know which values must be entered for ChangeFPS)
my avisynth script:
# I'm using the modded dll
SetMTMode(1,0)
# some sharpening, good results with good speed
asharp(1,6,-1,true)
# telecine to 59.94 fps: it creates a 3232.. pattern from 23.976 fps
ChangeFPS("ntsc_double")

I doubled the default values in Haali's renderer too:
Buffers: 65M , Frames: 82

Without Leak's patch: it stutters like hell.
But with it, sometimes it skips frames, maybe my comp is slow for it ...
My comp is: Athlon X2 4600, 1GB DDRAM, GeForce 7600GT (256MB)

And last: thanks guys for your hard work!

[Leak]

Originally Posted by chros

Is it makes sense at all to telecine 23.976 fps progressive content to 59.94 fps progressive one on a LCD TV/monitor running 60 Hz (from my comp) ?

Errr... if you really were to 3:2 telecine progressive video you'd actually get combed frames, I bet that's not what you want...

Anyway setting the buffers to 30/30 is really close to madness; the higher you set them the more lag you'll get and I'd not be surprised if your dropped frames come from that.

You *should* see the current best values for ahead and back in the OSD ("AviSynth info"), but you might not since there's something funny going on with it currently that I can't pinpoint...

Anyhow, I bet 0/1 (or 0 and a very small number) should be plenty for your use case, but honestly I think your best bet would be to just play your video as-is and let the renderer duplicate frames where neccessary.

Of course, setting your monitor to 72 Hz would be best...

np: Mouse On Mars - Schlecktron (Double Mix) (Rost Pocks)

[fastplayer]

@all SF members:
Don't forget to vote for ffdshow tryouts!

[mpioner]

Developers
plz add silent install options to InnoSetup

OK, cheers for that.

[Delerue]

I think that MPEG1 (both libavcodec and libmpeg2) decode is broken. I tested with this sample and noticed that it stutters a lot (at 0:11 sec, per example), and the aspect ratio isn't correct. The file was verified with MPC internal filter, Windows XP default decoder (with MPC and WMP11), VLC and SMPlayer (MPlayer), and it seems to run very well.

I tried with the build 1245 (inside the 'About' tab is written 1240, hehehe).

[B.F.]

Strange bug in SVQ3 decoder.
(Link was deleted)
It play fine with QuickTime decoder.

[chros]

Originally Posted by Leak

Errr... if you really were to 3:2 telecine progressive video you'd actually get combed frames, I bet that's not what you want...

Are you sure it's working that way?
From the manual: ChangeFPS changes the frame rate by deleting or duplicating frames. So I'm think it doesn't produces combed frames.

Originally Posted by Leak

Anyway setting the buffers to 30/30 is really close to madness; the higher you set them the more lag you'll get and I'd not be surprised if your dropped frames come from that. ...
Anyhow, I bet 0/1 (or 0 and a very small number) should be plenty for your use case, ...

Thanks. 0/1 values working pretty much the same like 30/30
ChangeFPS("ntsc_double") is still unwatchable, I'll try ChangeFPS("ntsc_video") which results a 1111211112... pattern.

Originally Posted by Leak

but honestly I think your best bet would be to just play your video as-is and let the renderer duplicate frames where neccessary.

If it's working this way, then probably this is the best choice. I'll ask Haali ...

Originally Posted by Leak

Of course, setting your monitor to 72 Hz would be best...

Yep, but my TV is only capable of 60 Hz.

[Leak]

Originally Posted by chros

Are you sure it's working that way?
From the manual: ChangeFPS changes the frame rate by deleting or duplicating frames. So I'm think it doesn't produces combed frames.

Ummm... what I meant was that ChangeFPS alone in no way, shape or form will ever produce "true" 3:2 pulldown, as 3:2 pulldown needs to be applied to fields, not frames.

Thanks. 0/1 values working pretty much the same like 30/30
ChangeFPS("ntsc_double") is still unwatchable, I'll try ChangeFPS("ntsc_video") which results a 1111211112... pattern.

What CPU load does the Windows task manager report? If it's around 50% on a dual-core machine it's probably too CPU-intensive... but seeing as it does work on a P4 2.8GHz here (okay, just the ChangeFPS("ntsc_double") part) I guess it's either the Asharp part gobbling up too much CPU or perhaps the usage of SetMTMode...

Yep, but my TV is only capable of 60 Hz.

Whoever is responsible for all kinds of displays being locked to 60Hz (often even in 50Hz PAL land) needs to be dragged out and shot with a cathode ray tube...

[foxyshadis]

Originally Posted by B.F.

Strange bug in SVQ3 decoder.
It play fine with QuickTime decoder.

SVQ3 with b-frames has had that problem for a while. It definitely sucks, I didn't find the problem last time but I'll look again - it might be a splitter issue though. (Sourceforge's tracker is impossibly bad for searching and sorting, but I think it isn't officially reported there. I'll take care of it. I'd donate if only they'd improve that and the mailing list, but I almost want to install Jira or Trac on the webspace instead.)

Bringing up a cam here was a bad idea though. Struck for rule 6.

[foxyshadis]

Put the changefps at the end of the script definitely, and try not to use other ffdshow filters after it. You'll now be (uselessly) filtering everything at 60fps instead of 24. You could also use convertfps (blended instead of jerky). Or make a short function to insert between each cycle a blend of frames 4 & 1 of the next, but that'll only be 29.97fps. I'd probably get too fancy for your tastes, I <3 the script.

Sometimes I really just wish you could have multiple avisynth instances, to run things like this after everything else.

[B.F.]

Originally Posted by foxyshadis

SVQ3 with b-frames has had that problem for a while. It definitely sucks, I didn't find the problem last time but I'll look again - it might be a splitter issue though. (Sourceforge's tracker is impossibly bad for searching and sorting, but I think it isn't officially reported there. I'll take care of it. I'd donate if only they'd improve that and the mailing list, but I almost want to install Jira or Trac on the webspace instead.)

Bringing up a cam here was a bad idea though. Struck for rule 6.

That was a smallest file I can find.
It was traler from one of movie news sites.
I just change extension from mov to mp4, so it can be played using Haali Splitter instead a QT.

P.S.
Can you delete that link from your post too?

[Leak]

Originally Posted by foxyshadis

Sometimes I really just wish you could have multiple avisynth instances, to run things like this after everything else.

*drops 10 foot pole*
*backs away slowly, trying to make no noise*
*turns around quietly*
*runs like hell*

[clsid]

Originally Posted by mpioner

Developers
plz add silent install options to InnoSetup

All InnoSetup installers are already capable of performing silent installs...

installer.exe /verysilent

(changing the options in the installer is possible too with /components="item1,item2,etc" /tasks="item1,item2,etc")

[chros]

Originally Posted by Leak

Ummm... what I meant was that ChangeFPS alone in no way, shape or form will ever produce "true" 3:2 pulldown, as 3:2 pulldown needs to be applied to fields, not frames.

So it was a little misunderstanding, caused by my not perfect knowledge

Originally Posted by Leak

What CPU load does the Windows task manager report? If it's around 50% on a dual-core machine it's probably too CPU-intensive... but seeing as it does work on a P4 2.8GHz here (okay, just the ChangeFPS("ntsc_double") part) I guess it's either the Asharp part gobbling up too much CPU or perhaps the usage of SetMTMode...

But with what content? I tried with 720p-AVC, and yes: one of my core is at 80%. I have tried with a standard xvid content (640*xxx) and it worked ... Thanks

Originally Posted by Leak

Whoever is responsible for all kinds of displays being locked to 60Hz (often even in 50Hz PAL land) needs to be dragged out and shot with a cathode ray tube...

At least it isn't interlaced ...

And an interesting thing with MPC-ffdshow_h264decoding on a dualcore CPU (no further filters are applied with the test):
1. start a 720p-AVC content, and see the loads of CPU (at here, core1 is minimal, core2 is very high)
2. stop the playback (don't close MPC), then start the playback again, and see the CPU: at here, the 2 cores is almost the same.
Can somebody reproduce this?
(I haven't installed AMD's DualCore Optimizer, but I use RMClock for it.)