FFAStrans forum

Hello,

I've been facing some issues with this specific type of files. The users send a tv show to some platform that uses AI to do audio translation. The file that they send back always stops during conversion. I have tried a lot of things and they always stop.
Here is the one thing I found weird in Media Info, and here is the log
In this workflow I was applying the normalization, but even removing everything. The job stops at the same place.

Can you guys maybe give me a light ?

Well, thats a really short last frame
In your log, i don't see that it did actually encode or deliver anything, maybe i am blind tough.

As this is not reproduceable independently of source file, we'll need some file sample from you to investigate.
But even without such a sample i guess i can tell that the issue is caused by some speciality of the source file. As aac "frame duration" is fixed to 1024 samples, it is just natural that the last frame will alwas be special. So with aac, the audio is either longer or shorter than the video but it will most likely never match the video duration (independent of the input file).
However, it is the first time that i see a negative frame duration (does not mean much, i am specialized in production encoding not delivery).

As a workaround:
My guess is that it works for you if you encode to mxf first, no matter which codec, because mxf automatically forces exact same audio and video duration. I would try this first just to get it working once. Moving on i would play with different "pre rewrappings", maybe rewrap to mov but choose audio codec pcm_s24le.
Example commandline, not sure if this works: And set cmd processor to set s_source variable to
without quotes

After that, analyze the loudness and encode (or do this after analyzing but before encoding, as you prefer)

Thanks emcodem! I will try these options!
The job failed at the encoding node, it never went to the deliver node.
I could deliver as a MXF, if that solves it. I'll just change to that deliver node!

I'll let you know how it goes!

Just to let you know, as we didnt find the needed hints for what was going wrong in your job json,
Steinar added also some more logging for the next version. Not helpful for you but just fyi

Hey guys! I tried the command executor way and using all different types of MXF, everything fails.
I tried decoding A/V and see if could bypass the audio, I was thinking of a way to split audio and video, and mix them together, to see if that could help out, but dind't find a way of doing that. But it will probably fail also...

Is there a way for you to provide a source file sample?

Unfortunately even using that file i cannot reproduce any issue, even when using the workflow you sent.
You sure the issue happens with this very file too?

Can you run a decoding check with ffmpeg on one file that does not work for you? Also, as we see exiftool errors in your workflow aboe, can you check if exiftool works when used manually?

Hey emcodem,

Still nothing.
Here are some logs of the attempts, using the exact same file I sent you.

OK the issue must be completely different than i thought. What fails is a normal exiftool analysis but as it works on my installations (tried different ffastrans versions), it should not be about a source file issue nor a software issue.

Can you try if this commandline works on MANVOLUMES?

C:\FFAStrans\processors\ffmpeg\exiftool.exe -ee \\172.168.100.20\Transfer\ENTRADA\VTS_HUB\pt_Migrant_Workers_Explains_Dotcom_Final_translated.mp4 -api largefilesupport=1 -json -charset filename=UTF8