FFMS2000 - The experimental future of FFMS2

[Myrsloik]

Originally Posted by sneaker_ger

Yes, looking good.

Don't forget to let test5 create new .ffindex files, people.

Yes, delete all old ones. I forgot to bump the index version and using an old index=>old issues.

[dipje]

The weird things I had at the start of my MTS files seem fixed now. Stepping through the files (forward) reveal nothing out of the ordinary an all good.

Stepping backwards still starts acting weird, but you might just call that 'by design and do not seek backwards' I don't know.
If I start stepping backwards (after having stepped forward through the whole video) the first few frames seek very quick (cached I guess) then the frames seek back slower (to be expected, seeking backwards in a file that is not All-I).
But after around 80 frames the picture 'freezes'. I can step back a few frames but they all look the same (they didn't while seeking forward through the file). This lasts for around 10 frames (GOP length?). Then it goes to slowly seeking backwards frame by frame again.

Now, if I start stepping forward at this moment, those 10 'freeze frames' are still there. I have to step back a whole lot more and then step forward ago and now the exact same frames step forward OK. So is there actually some seeking gone bad here, or is the Vapoursynth cache being naughty? Or both?

What appears to be happening is that after seeking +/- 80 frames backwards ffms2 produces 10 frames of 'freeze frame', and Vapoursynth caches those 10 frames, so stepping them forward again will not fix it while the bad frames are still in the cache.

@Myrsloik: You still have my MTS example file right? Create a simple script that opens it with ffms2 with nothing more, open it in Vapoursynth Editor. Go to the last frame, then start holding down the arrow-down key to step backward. Around frame 192 the 'freeze frame' occurs. Also happens in VDFilterMod (to rule out a bug in Vapoursynth Editor or something).
If I open the file and don't seek to the end but instead seek to frame +/- 220 and start holding down the arrow key, the glitch once again occurs at frame +/- 192 (same spot) so it does seem related to that position in the file, not the amount of frames I stepped backwards or anything.

PS (And yes, I deleted .ffindex file before I started)
PS2 (And I'm talking about test5, just to be clear)

[stax76]

Originally Posted by Myrsloik

It would be very helpful if you throw all your mpeg and transport streams at it.

How would I handle audio, dump or audiodub?

I want to encode this sample:

http://www.mediafire.com/file/t51xbk...owing_Pains.ts

Some script code and command lines to get me started would be helpful.

[Myrsloik]

Originally Posted by dipje

The weird things I had at the start of my MTS files seem fixed now. Stepping through the files (forward) reveal nothing out of the ordinary an all good.

...

Interesting. It at least seems to improve and probably exposes a few different issues. So many old hacks to re-evaluate in there...

I wish I'd kept a bigger collection of samples.

[Atak_Snajpera]

Is this normal that in latest ffms2000 test5 all frames in VC-1 are detected as keyframes?
Example

# keyframe format v1
fps 0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

FFMS 2.20

# keyframe format v1
fps 0
0
1
25
49
53
77
101
106
130
154
178
202
226
250
274
298
322
346
370
394
418
442
466
490

[Myrsloik]

Originally Posted by Atak_Snajpera

Is this normal that in latest ffms2000 test5 all frames in VC-1 are detected as keyframes?
...

No, can I have a small sample?

[Atak_Snajpera]

No problem give me 15 minutes...

[Atak_Snajpera]

ok done
http://www.mediafire.com/file/98z5ynb5m6dwb0r/VC-1.mkv

[Myrsloik]

Originally Posted by Atak_Snajpera

ok done
http://www.mediafire.com/file/98z5ynb5m6dwb0r/VC-1.mkv

Must be a broken mkv/ffmpeg bug. The parser really does report all the frames as keyframes. I'll poke it a bit more but definitely not my fault.

[FranceBB]

It seems that the latest ffms2000 breaks BePipe -> NeroAACEnc compatibility.
Although the audio stream is detected by BePipe, NeroAACEnc can't encode it and fails.

[Myrsloik]

Originally Posted by FranceBB

It seems that the latest ffms2000 breaks BePipe -> NeroAACEnc compatibility.
Although the audio stream is detected by BePipe, NeroAACEnc can't encode it and fails.

Audio handling wasn't changed at all. Oddly enough.

[FranceBB]

Uhm... Well, with the stable FFMpegSource2 I was able to open a file in avisynth, and pipe the audio using BePipe to NeroAACEnc, but I can't with ffms2000.
I think you can reproduce my error pretty easily with any source.

Sample:

FFMpegSource2("Ep1.avi", fpsnum=24000, fpsden=1001, atrack=-1)

BePipe.exe --script "Import(^AVS Script.avs^)" | neroAacEnc.exe -lc -br 320000 -if - -of "audio.m4a"

[FranceBB]

Hi,
I have an .mka audio track which contains two tracks: the first one is a FLAC lossless audio track with 6 channels, and the second one is an AC3 stereo audio track.
When I use FFAudioSource() it takes the second track (AC3) instead of the very first one.
I think it should use the first available audio track, so FLAC.
Is this a bug? Do you want me to upload the .mka file?


General
Unique ID : 174555563715229058114849233714037036678 (0x83522DC00F66393A809D00CAF83E9686)
Complete name : O:\episodi\[VCB-Studio] Shingeki no Bahamut Genesis [Ma10p_1080p]\bahamutep2audiodolby.mka
Format : Matroska
Format version : Version 4 / Version 2
File size : 664 MiB
Duration : 23 min 52 s
Overall bit rate mode : Variable
Overall bit rate : 3 890 kb/s
Encoded date : UTC 2016-11-12 14:21:14
Writing application : mkvmerge v9.5.0 ('Quiet Fire') 64bit
Writing library : libebml v1.3.4 + libmatroska v1.4.5

Audio #1
ID : 1
Format : FLAC
Format/Info : Free Lossless Audio Codec
Codec ID : A_FLAC
Duration : 23 min 52 s
Bit rate mode : Variable
Bit rate : 3 249 kb/s
Channel(s) : 6 channels
Channel positions : Front: L C R, Side: L R, LFE
Sampling rate : 48.0 kHz
Frame rate : 11.719 FPS (4096 spf)
Bit depth : 24 bits
Stream size : 555 MiB (84%)
Writing library : libFLAC 1.2.1 (UTC 2007-09-17)
Language : Japanese
Default : Yes
Forced : No

Audio #2
ID : 2
Format : AC-3
Format/Info : Audio Coding 3
Format settings, Endianness : Big
Codec ID : A_AC3
Duration : 23 min 52 s
Bit rate mode : Constant
Bit rate : 640 kb/s
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 48.0 kHz
Frame rate : 31.250 FPS (1536 spf)
Bit depth : 16 bits
Compression mode : Lossy
Stream size : 109 MiB (16%)
Language : Japanese
Service kind : Complete Main
Default : No
Forced : No

[Myrsloik]

Are both tracks manually selectable?

[FranceBB]

Yes, sure. I can select them using potplayer. If I just open the file, it plays the FLAC track as default.

[Myrsloik]

Released test6. Download link in first post.

[Mystery Keeper]

Test6 finally opens VP9 properly. Thank you!

[stax76]

@Myrsloik

I believe there is a memory leak with the x64 build, I've not examined it in detail, in staxrip for every job memory grew 50-100 MB, I always thought it's staxrip and I've been too careless with resources, forgetting to detach event handlers and things, I changed my design and had major problems with it, by accident I noticed that it don't happen with l-smash, I've tried with VirtualDub, after every opening and closing circle memory grows 10 MB, for GUIs this is very problematic because they don't keep one script open but rather work with multiple scripts opening and closing them on demand. I've currently an issue on the tracker by somebody that is trying to batch process 3200 files in one go. :-)

[manolito]

That's what you get from including experimental versions of helper applications in a stable version of your own software...

[stax76]

I knew it was coming.