Audio out of sync when using ffmpeg, adelay, and amix

[AspiringVideoProcessor]

Good Afternoon,

I have several webm files. I also have startTimeOffset and stopTimeOffset of the videos.

Offsets defined as:

• startTimeOffset — The offset, in milliseconds, for when this file started recording (from the createdAt time for the archive).

• stopTimeOffset — The offset, in milliseconds, for when this file stopped recording (from the createdAt time for the archive).


Here is an example:

```
d4a19a58-790a-45c1-b96e-819bc98cd408.webm | 1720 | 3762173
ab794f92-bdfb-4bb0-821d-ba80c8a9228a.webm | 48920 | 3328385
a7df9a51-0671-4930-a14b-e7d85a562577.webm | 76220 | 3763361
5d312878-d25d-40d7-9418-bed85e22efc6.webm | 3109640 | 3267473
1cf6127d-250a-4601-a219-b1b6bbd725f8.webm | 3360980 | 3636927
c3ddd462-e05a-425f-af58-f288f930f2cb.webm | 3689600 | 3760423

```

These files are indivual users in a group conversation.


My end goal here is to combine them all into 1 audio file

Here is what I have done so far. It creates the group audio just fine but the audio isn't lined up. there is like a 33 second delay from when person 1 talks and then person 2 answers. Then the rest of the video is off and on ocassion people are talking over each other because the audio is out of alignment.


Here is the command I used to get the webm file to convert to MP3 broken into lines for readability:

```

ffmpeg
-acodec libopus
-i d4a19a58-790a-45c1-b96e-819bc98cd408.webm
-b:a 320K
-vn
-y d4a19a58-790a-45c1-b96e-819bc98cd408.mp3

```

Here is the full command I used to combine all the MP3s with the delays.

```
ffmpeg
-i d4a19a58-790a-45c1-b96e-819bc98cd408.mp3
-i ab794f92-bdfb-4bb0-821d-ba80c8a9228a.mp3
-i a7df9a51-0671-4930-a14b-e7d85a562577.mp3
-i 1cf6127d-250a-4601-a219-b1b6bbd725f8.mp3
-i c3ddd462-e05a-425f-af58-f288f930f2cb.mp3
-filter_complex "
[0]adelay=1720|1720[a0];
[1]adelay=48920|48920[a1];
[2]adelay=76220|76220[a2];
[3]adelay=3360980|3360980[a3];
[4]adelay=3689600|3689600[a4];
[a0][a1][a2][a3][a4]amix=inputs=5
"
out.mp3 -y
```

As mentioned if I take like 33 seconds off of the 76220 offset, person 2 ends up answering the question at the correct spot in the video.

There is no errors in the output. Everything appears to process correctly.


The Mp3 files are different durations. For example, person 1 was on the conference for 1 hour, person 2 might only be on the call for a half hour, though in this particular instance person 1 was on the call for 74.5 seconds before person 2 connected.


All the examples I have read seem similar to what I am trying to do.

Any ideas?

[ProWo]

My end goal here is to combine them all into 1 audio file

Do you want to create only one continuous audio track or do you want to synchronize a continuous audio track with a video track?
In the first case, audio delays are irrelevant.
Try to explain your intention better.

[davexnet]

I don't understand why you would even use FFmpeg when an audio editor, Audacity for example,
is much better suited

[AspiringVideoProcessor]

Think of a video zoom call.

You have 4 people all the call.

They are all conversing in a group meeting.

When the call over. I am given a zip file with 4 webm files.

Those 4 files are individual streams.

I am tasked with combining those streams into 1 combined stream so people can listen to the audio file and hear the conversation.

I have a situation where the 4 users joined the video call but did not enable their video for privacy concerns.

The files still come over as webm with different start & end time offsets.

I convert the webm files to mp3.

I need to combine all the mp3 files using the offsets so I can play back the original conversation with all the audio lined up so the conversation makes sense.

[ProWo]

So you have one audio track per person, on which only their voice can be heard?
Then you can load these individual tracks together into a multitrack audio editor and then move the individual tracks around until they form a fluid conversation. Audacity can do this, but you should not convert the webm to mp3, but to pcm(wav), then merge and edit this and then convert the result to mp3.

[richardpl]

To the original poster: always use latest git master build.

Also to others posters, topic is about ffmpeg and not about other less useful software.

[AspiringVideoProcessor]

For those that are wondering why I am using ffmpeg instead of another tool, here is some more information:

1. Most of the archives are webm video and audio together. I use ffmpeg to convert to mp4 and then do a grid video (2x2,3x3,4x4) of these meetings.

2. Automation. Because there are too many videos to do this one meeting at a time.


The issue in my original post came about when recently they decided to do video interviews because they need to share the screen but did not turn on their videos.

I speculate that something happened with the start time offsets, but I am not sure. It could be that I am missing something on how things work.

I will upgrade to the latest git version today to see if that helps.

[AspiringVideoProcessor]

Ok, good news, after using a ffmpeg static build based on master, the above command worked fine. Thank you for the suggestion, @richardpl.