'atempo' didn't work unless '-map 0' was added

[markfilipak]

source.vob is here: ​https://www.dropbox.com/s/bdahs4n08z120z6/source.vob?dl=0

To reproduce the issue, run 1.cmd .. 5.cmd. That will take less than a minute. Then read the rest of this report.

1.cmd => ffmpeg -i source.vob -filter_complex "[0:v]settb=expr=1/24,setpts=expr=N[v]" -map "[v]" -map 0:1 -codec:v %AVcoder% -codec:a copy -r 24 1.mkv

2.cmd => ffmpeg -i source.vob -filter_complex "[0:v]settb=expr=1/24,setpts=expr=N[v];[0:a]atempo=1.001[a]" -map "[v]" -map "[a]" -codec:v %AVcoder% -codec:a ac3 -r 24 2.mkv

3.cmd => ffmpeg -i source.vob -filter_complex "[0:v]settb=expr=1/24,setpts=expr=N[v];[0:a]atempo=1.001[a]" -map 0 -map "[v]" -map "[a]" -codec:v %AVcoder% -codec:a ac3 -r 24 3.mkv

4.cmd => ffmpeg -i 3.mkv -map 0:0 -map 0:1 -codec:v copy -codec:a copy 4.mkv

5.cmd => ffmpeg -i 3.mkv -map 0:2 -map 0:3 -codec:v copy -codec:a copy 5.mkv

source.vob has a 9-&-1/2 second clip that's repeated 7 more times. The clip is soft telecined. And of course, the video and audio are synchronized.
1.mkv, 2.mkv, 3.mkv were made by converting the soft telecined source to 24fps.
4.mkv, 5.mkv, were made from 3.mkv.

1.mkv was made just to show that the audio lags the video at the end of the 8th clip. Note the boy slamming the door. By the end of the 8th clip, it lags by about 1/2 second.
2.mkv is what I expected would produce synchronized audio, but it fails.
3.mkv appears to work when played with MPV, but there's an extra video track and an extra audio track.

4.mkv is just the 1st video & audio pair of tracks from 3.mkv.
5.mkv is just the 2nd video & audio pair.

4.mkv is what I want, so output tracks 0:0 & 0:1 must be from the graph.
5.mkv is not what I want, so output tracks 0:2 & 0:3 must be from '-map 0'.

Issue #1: Why does 'atempo' need '-map 0' to work?
Issue #2: Will what I want always be 0:0 & 0:1?

1.mkv, 2.mkv, 3.mkv, 4.mkv, 5.mkv are available on request.
1.log, 2.log, 3.log, 4.log, 5.log are available on request.

Regards, And Thank You for ffmpeg.
Mark Filipak.

[Gyan]

-map 0 -map "[v]" -map "[a]" will first include the source (unmodified) tracks and then the filtered tracks. Check the stream mapping in the log of 3.cmd

[markfilipak]

Replying to Gyan:

-map 0 -map "[v]" -map "[a]" will first include the source (unmodified) tracks and then the filtered tracks. Check the stream mapping in the log of 3.cmd

From 3.log, you can't tell which output tracks are from the graph and which tracks are from '-map 0'.

[markfilipak]

Sorry, I guess replying to the email doesn't work. So, I'll repeat my reply here, now.

Hi Gyan, Thanks.

Actually, it's the other way around; [0:0] & [0:1] are from the graph and [0:3] & [0:4] are from '-map 0'. Yes, that was surprising, but it's not really the issue.

Do you want 1.mkv..5.mkv?
With the source -- link provided -- and the scripts,
you can generate 1.mkv..5.mkv in less than a minute,
but I'll attach them if you prefer.

The issue is that without '-map 0', 'atempo' doesn't function. That can be seen by watching the last 10 seconds of 2.mkv (without '-map 0') v. the last 10 seconds of 3.mkv (with '-map 0').

3.cmd fixes that -- see 3.mkv -- but the '-map 0' adds the extra tracks.

4.mkv & 5.mkv show which tracks were added by the graph and which were added by '-map 0'.

• Mark.

[Gyan]

Replying to markfilipak:

Sorry, I guess replying to the email doesn't work. So, I'll repeat my reply here, now.

Hi Gyan, Thanks.

Actually, it's the other way around; [0:0] & [0:1] are from the graph and [0:3] & [0:4] are from '-map 0'. Yes, that was surprising, but it's not really the issue.

Check again. From 3.log

`
Stream #0:0 -> #0:0 (mpeg2video (native) -> hevc (libx265))
Stream #0:1 -> #0:1 (ac3 (native) -> ac3 (native))
setpts (graph 0) -> Stream #0:2 (libx265)
atempo (graph 0) -> Stream #0:3 (ac3)
`

[markfilipak]

Replying to Gyan:

Replying to markfilipak:

Sorry, I guess replying to the email doesn't work. So, I'll repeat my reply here, now.

Hi Gyan, Thanks.

Actually, it's the other way around; [0:0] & [0:1] are from the graph and [0:3] & [0:4] are from '-map 0'. Yes, that was surprising, but it's not really the issue.

Check again. From 3.log

Stream #0:0 -> #0:0 (mpeg2video (native) -> hevc (libx265))
Stream #0:1 -> #0:1 (ac3 (native) -> ac3 (native))
setpts (graph 0) -> Stream #0:2 (libx265)
atempo (graph 0) -> Stream #0:3 (ac3)

A sidetrack: Ah! You're right. Sorry. But something is strange. 4.mkv is:

Stream mapping:
  Stream #0:0 -> #0:0 (copy)
  Stream #0:1 -> #0:1 (copy)

and has synchonized audio as shown when the boy slams the screen door (so 'atempo' in the graph worked for them), while 5.mkv is:

Stream mapping:
  Stream #0:2 -> #0:0 (copy)
  Stream #0:3 -> #0:1 (copy)

(so are from '-map 0').
Something is backwards.

To the real issue, why does 'atempo' not work unless I add '-map 0'?

[markfilipak]

I've given it some thought. Since 4.mkv & 5.mkv are made from 3.mkv, 3.log must be showing the mapping wrongly. It appears there's a second problem there.

-EDIT- On third thought, maybe 4.cmd & 5.cmd are not mapping properly, eh? I see the mapping as follows:

3.mkv...........4.mkv..5.mkv
graph -> 0:0 -> 0:0
graph -> 0:1 -> 0:1          door slam is correct ('atempo' worked)
input -> 0:2 --------> 0:0
input -> 0:3 --------> 0:1   door slam is late

[markfilipak]

I spent the last week creating a very complicated Windows script, but now I'm ready to follow-up on this ticket.
In the last few minutes I altered the script to save the intermediate (temp) file (before stripping out the duplicate streams) that has all the original streams, doubled.
You are correct and I was wrong: The 1st set of streams are from '-map 0' and the 2nd set of streams are the ones added by '-filter_complex "[0:v]settb=expr=1/24,setpts=expr=N[v];[0:a]atempo=1.001[a]" -map "[v]" -map "[a]"' -- I used MKVToolNixGUI to figure it out, saving the lower numbered streams to one output and the higher numbered streams to another output.

I don't know how my mistake happened and I appolgize for it -- having a script should avoid such a mistake in the future.

I'm going to do more testing with more movies to see whether the sets of streams (one from '-map 0' and the other from the '-filter_complex') remain reproduceable.

The bottom line remains: Without '-map 0' in the command, 'atempo' does not take affect.

Should I prepare an entire new presentation and ticket, or should I append a new presentation to this ticket? Your call.

Regards,
-Mark.

[Elon Musk]

Needs simplified report.