Hybrid(Windows/Linux/Mac): Input -> x264/x265/Xvid/VP8/VP9

[Selur]

Thanks for the info.
WebSpace provide again messed up the LetsEncrypt update,...
(I reported the issue, hopefully this will be fixed fast,..)

Cu Selur

[sophisticles]

Is there a reason why the Linux builds no longer support qsvenc, or any hardware encoder for that matter?

I could swear that they used to, but the last few releases have been missing them and a look through the folder shows that the binaries are not included.

[Selur]

The Linux build never supported NVEncC, QSVEncC,....

[firest]

Please, would it be possible to add other reader options for QSVEnc? Apparently currently it's only possible to use --y4m or --avhw. The reason I ask this is because other options like --avs or --vpy work much faster on my computer, maybe because they require less resources from it...

Thanks for making this great program.

[Selur]

From my tests so far for neither NVEncC/QSVEnC/VCEnC it's faster to use '--avs' or '--vpy' instead of using ffmpeg or vpsipe and pipe the output as y4m to the encoder.
If can you share some examples where it's faster (and I can reproduce this) to use '--avs' or '--vpy' I will look into writing support for it.

Cu Selur

Ps.: looked at it, I know how to implement it, but the question is whether it's worth it. Atm. I suspect, that especially more complex scripts might even be faster if decoded outside of encoder on systems with 4+ cores.

[firest]

Since I use a low end computer (4 cores processor, with a base frequency of 2.0 GHz), you will probably never really be able to reproduce.
I don't know exactly why, but according to my tests I gain around 20 fps during the encode (around 66% improvement) leaving the default value of QSVEnc, which in this case will pick one depending on the extension of input file (in my case it's usually .avs or .vpy). But I understand if it is too much work and you don't want to invest your time in it.

[Selur]

Send you a pm.

Cu Selur

[firest]

Just tested the dev version you sent me, and it works fine for ".vpy" extension, but not for ".avs". Apparently it sets "--vpy" in either case.
Also, the option to replace "--y4m" will only work in cases where AviSynth or VapourSynth are actually being used, when using some filter, for example. This is easily worked around by enabling "Always use Avisynth" in "Config -> Internals -> Avisynth", however there is no similar option for VapourSynth, I guess. That said, in these cases "--y4m" will be used even with "DirectSynthLoading" enabled.
So I think a better approach would be to remove "--y4m" from the command line when "DirectSynthLoading" is enabled, and just let QSVEnc select one automatically depending on the extension of input file. Or let the user choose one manually through "QSVEnc -> Misc -> Custom command line addition". Only suggestions, it's up to you.

In any case, thanks for taking some of your time to look at this. I'm leaving a debug output in case you want to take a look.

[Selur]

Ah, forgot that Avisynth/Vapoursynth are actually needed, not that they are selected in support, will fix.
Will look at it today after work.

Had a quick look at the debug output, that does not seem to be the issue.

2023.05.09 - 19:03:47_Windows 10 Version 21H2 (64bit)_2023.05.09.1 - level 9: QSVENC output: vpy: VapourSynth script error.
vpy: Python exception: (unicode error) 'unicodeescape' codec can't decode bytes in position 2-3: truncated \UXXXXXXXX escape (<undefined>, line 7)
vpy: Traceback (most recent call last):
vpy:   File "src\cython\vapoursynth.pyx", line 2852, in vapoursynth._vpy_evaluate
vpy:   File "<undefined>", line 7
vpy:     LSMASHVideoSource("C:\Users\Public\Videos\nature.mp4",format="YUV420P8")
vpy:                                                          ^
vpy: SyntaxError: (unicode error) 'unicodeescape' codec can't decode bytes in position 2-3: truncated \UXXXXXXXX escape
failed to initialize file reader(s).

-> Seems like QSVEnc can't decode some special character in the Vapoursynth script. (works fine here :/)
Can you share the .vpy file you are using (the original from the file system) with me?

Ahh, that .vpy script isn't created by Hybrid and wrong.

Either use slashes:

LSMASHVideoSource("C:/Users/Public/Videos/nature.mp4",format="YUV420P8")

or add an 'r' before the String

LSMASHVideoSource(r"C:/Users/Public/Videos/nature.mp4",format="YUV420P8")

-> this isn't a bug in Hybrid

Cu Selur

[firest]

Actually, there was no ".vpy" script, but rather an ".avs" one.
As I mentioned before, it sets "--vpy" even for the ".avs" extension, so probably this is the problem. It tries to read the ".avs" script as if it were ".vpy", hence the problem with the slashes, absence of "r" or something else.
And yes, the script was created by Hybrid.

[Selur]

Send you a new link.

Cu Selur

[firest]

Hey, just tested the new dev version and it is now setting "--avs" and "--vpy" correctly. Great!
The only trade-off I could notice is that if you are not using "--y4m" but "--avs" or "--vpy", the estimated time to finish the encode doesn't show up in the progress bar.

Also, as I mentioned before:

the option to replace "--y4m" will only work in cases where AviSynth or VapourSynth are actually being used, when using some filter, for example. This is easily worked around by enabling "Always use Avisynth" in "Config -> Internals -> Avisynth", however there is no similar option for VapourSynth, I guess. That said, in these cases "--y4m" will be used even with "DirectSynthLoading" enabled.

So, to be able to use "--vpy" it is necessary to actually be using VapourSynth in some way, since for the parameter to work it requires a ".vpy" script defined as input file. Since there is no way to always force its use (as is possible with AviSynth), "--y4m" will be used even if "DirectSynthLoading" is enabled. If it's not too much work, maybe you could introduce the same "Always use avisynth" functionality for VapourSynth as well.

Anyway, thanks a lot for your support!

[Selur]

the estimated time to finish the encode doesn't show up in the progress bar.

will look into it (having 12hr workdays atm. to not much time )

maybe you could introduce the same "Always use avisynth" functionality for VapourSynth as well.

already there "Filtering->Vapoursynth->Misc->Always use Vapoursynth"

Cu Selur

[firest]

will look into it (having 12hr workdays atm. to not much time )

No problem, take your time.

already there "Filtering->Vapoursynth->Misc->Always use Vapoursynth"

Oh, I didn't realize that. Thanks for the info.

[Selur]

Send you a link to a dev version which should fix the progress indication.

Cu Selur

[firest]

Yeah, now it's fixed. Thanks a lot

I have one last question: is it possible to pass extra parameters to the command line of audio encoders, as is possible for video encoders? (eg: QSVEnc -> Misc -> Custom command line addition)

[Selur]

atm. there is no functionality for that.
What parameters do you want to add to which encoder?

Cu Selur

[firest]

What parameters do you want to add to which encoder?

Well... I noticed that there are several missing, this is the reason why I asked about the custom command line, but I can name a few:
For qaac, for example, I quite often use "--start"/"--delay". I particularly prefer to add the delay here during encoding rather than leaving it to the video muxer (except when there is no need to encode/re-encode the audio). I also like "--normalize" (or "-N"), as there is no need to specify a dB value and you can still use "--gain" , if desired, for this. There are also various tags for music files, which I often use for audio in video clips.
For fdkaac possibility to turn on or off "--afterburn", be able to set a frequency bandwidth when using AAC LC with "--bandwidth", among others. So a custom command line would be great, but if a lot of work, "--start"/"--delay" and "--normalize" would be fine atm, for qaac, which is what I've been using most often.

P.S. Sorry if I'm asking too much , having fulfilled the previous request was already great.

[Selur]

regarding normalize: What is wrong with the normalization Hybrid offers through sox? (personally I prefer: Gain->auto + DRC->Film Standard)
using: start&delay: this only seems to make sense when processing audio only, otherwise it seems to be better to let Hybrid and the decoder do the trimming through the cut option.
regarding fkdaac afterburn and bandwidth: I will think about adding these in the ui.

Cu Selur

[firest]

regarding normalize: What is wrong with the normalization Hybrid offers through sox? (personally I prefer: Gain->auto + DRC->Film Standard)

Nothing wrong, I only talked about "--normalize" because I usually use it with qaac and always get good results, but I'ill try the way you mentioned.

using: start&delay: this only seems to make sense when processing audio only, otherwise it seems to be better to let Hybrid and the decoder do the trimming through the cut option.

Yes, it's for audio only. But now that you mention it I remembered a detail, using Hybrid is there any way to get an audio-only output format? eg.: .m4a, .aac, .ec3, .eac3, etc.

[Selur]

Yes, it's for audio only. But now that you mention it I remembered a detail, using Hybrid is there any way to get an audio-only output format? eg.: .m4a, .aac, .ec3, .eac3, etc.

Sure, set video to 'ignore' in case source isn't video only and set output to .raw Hybrid will then add the extension based on the format.

Cu Selur

[Selur]

Got a question regarding fkdaac afterburn:
Why do you want to disable it? (since it's on by default)

 -a, --afterburner <n>         Afterburner
                                 0: Off
                                 1: On(default)

Cu Selur

[firest]

Sure, set video to 'ignore' in case source isn't video only and set output to .raw Hybrid will then add the extension based on the format.

Perfect. Thanks.

Why do you want to disable it? (since it's on by default)

As mentioned in the man page:

   -a, --afterburner <n>
          Configure afterburner mode.  When enabled, quality is  increased
          at the expense of additional computational workload.

So on big input files it can take ages to finish the encode, since I use a low-end computer.


I also have a question about something I just noticed:
As you can see in the screenshot below, "--colormatrix" is added twice, is this intentional?

[Selur]

Don't see a screenshot,.. but I checked the code an in QSVEnc VUI setting are set twice, which shouldn't break anything, but I'll fix it.

Cu Selur

[firest]

Strange... to me it appears. But you found the right thing anyway.

[firest]

Hello, Selur. Could you send me the link to the latest dev version of Hybrid that you had sent me by pm (with fdkaac changes) again?
I hadn't seen it yesterday, and when I tried to download it today, apparently the link had expired (?). Thanks.

[Selur]

send you a link

[firest]

Got it, thanks.

Well, I was trying to encode an audio using qaac a few minutes ago and the job aborts with a message that says "CRASHED".
So I tried again to generate a debug file and see if I could find anything that would help me solve it on my own, but the only thing that seemed to indicate anything was this:

level 9: Main call started,..
level 9: Setting m_startingJobs(2) to FALSE
level 9: Qaac output: -o is not available for multiple output.
level 9: process finished with exitcode: 1 and exitStatus: 0

Any idea? I'm leaving the debug file below.

[Selur]

Looking at the encoding call:

ffmpeg -y -threads 2 -loglevel fatal -nostdin -i "F:\Videos\Temp\en_1.aac" -ac 2 -ar 48000 -f sox - |  sox --multi-threaded --temp "F:\Videos\Temp\2023-05-13@15_15_22_431000" --buffer 524288 -S -t sox - -b 16 -t raw - compand 0.1,0.3 -90,-90,-70,-64,-43,-37,-31,-31,-21,-21,0,-20 0 0 0.1 gain -n |  qaac --threading --raw --raw-channels 2 --raw-rate 48000 80 --adts --he - -o "F:\Videos\iId_7_aid_0_DELAY_-108ms_2023-05-13@15_15_22_431000.aac"

The '--raw-rate 48000 80' part is the issue.

It probably should be "--raw-rate 48000 --tvbr 80" - -o "J:\tmp\iId_1_aid_1_2023-05-13@20_56_49_7510_02.aac".
Seems like I broke bit rate encoding for aac during my last changes,..
Will look at it tomorrow.

Cu Selur

[Selur]

Send you a new link via pm.

Cu Selur