[FFmpeg-user] Concat demuxer avoid injecting additional silent packets

[Ronak]

> On Oct 11, 2022, at 1:45 AM, Erik Dobberkau <erik.dobberkau at gmail.com> wrote:
> 
>>> Is there a way to avoid having FFmpeg inject those additional silent
>> packets while concatenating?
>> 
>> Here is my FFmpeg information: […]
>> 
> 
> Is your audio encoded using AAC? Your uncut console output would tell had
> it not been omitted… therefore, please make sure to always provide the
> complete console output.
> 
> IIRC, AAC has packets of 1024 samples, and the concat (de)muxer/filter may
> operate on packet level, not sample level, which would explain the silence
> (missing samples to complete their number to the next integer multiple of
> 1024).
> 
> It’s only a guess though, and (one of) the developers should be able to
> share some insight on what’s (not) happening.

Yes my audio is encoded using AAC. From what I see in ISOViewer, the first file has 91410 samples with a sample duration of 1024 (what we’d expect). Then it has 1 sample with a duration of 456. 

When I look at the concatenated file, I see 91410 samples with that same 1024 sample duration. Then I see 1 more sample with a duration of 461.

I tried trimming this sample on my own before the concat, but even if I pass in the exact duration (91410 * 1024/<sample rate>) to the trim command, FFmpeg doesn’t actually remove that last sample.

Any tips on this?

I would attach the complete FFmpeg output, but it’s rather long….

> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
> 
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".