#9653 closed defect (fixed)
mka to mp3 but talk missing
Reported by: | Monika Hoffmann | Owned by: | |
---|---|---|---|
Priority: | normal | Component: | avcodec |
Version: | git-master | Keywords: | opus |
Cc: | Monika Hoffmann | Blocked By: | |
Blocking: | Reproduced by developer: | yes | |
Analyzed by developer: | no |
Description
Summary of the bug:
I converted 2 mka into mp3, user1.mp3 is ok, but in user2.mp3 the talk is missing. They alternately count '1 2 3', '4 5 6', ...
The internet connection is bad for User2
How could I use ffmpeg to get a mp3 with talk.
Using MKVToolNix I get a mp3 with talk, so there should be a way
How to reproduce:
% ffmpeg -i "user1.mka" -vn -ab 48k -ar 16000 -ac 2 -c:a libmp3lame "user1.mp3" % ffmpeg -i "user2.mka" -vn -ab 48k -ar 16000 -ac 2 -c:a libmp3lame "user2.mp3 ffmpeg version 2022-01-24-git-0a83ecbf48-full_build-www.gyan.dev Copyright (c) 2000-2022 the FFmpeg developers built with gcc 11.2.0 (Rev7, Built by MSYS2 project)
Attachments (5)
Change History (11)
by , 3 years ago
by , 3 years ago
by , 3 years ago
by , 3 years ago
comment:1 by , 3 years ago
comment:2 by , 3 years ago
I listened to the user2.mka again with "VLC Plus Player".
In the 10th second he speaks for the first time:
"Eins, äh vier fünf sechs"
17th sec: "zehn"
20th sec: "zehn elf zwölf"
28th sec: "sechzehn siebzehn achtzehn"
40th sec: "ok"
47th sec: "ok und jetzt sage ich tschüss, hörst du mich?"
53th sec: "alles klar, tschau"
In MKVToolNix GUI v65.0. - Multiplexer I add user2.mka to the queue and enter user2_toolnix.mp3 to the target, then start Multiplexer
I add the user2_toolnix.mp3
yes, you are right: They lack the audio sample rate and channel count fields in the header.
We get them like this.
My job is to write a program (in javaScript/node) to get an mp3 (or m4a) from mka and then merge them into one file where they talk to each other (together.mp3).
Sometimes also with user1.mkv as together.mp4
comment:3 by , 3 years ago
Component: | undetermined → avcodec |
---|---|
Keywords: | opus added |
Reproduced by developer: | set |
- Our native Opus decoder produces garbage with user2.mka; workaround is to use libopus for decoding (use "-c:a libopus -i user2.mka").
- MKVToolNix does not produce a real mp3; it produces a Matroska file with Opus audio in it. Using an mp3 extension does not make a proper mp3 out of it.
comment:5 by , 2 years ago
Resolution: | → fixed |
---|---|
Status: | new → closed |
Fixed in e301143f96639b97a8e9a0045f93b56e0f399289. (This makes #9890 basically a duplicate of this ticket.)
comment:6 by , 2 years ago
Priority: | important → normal |
---|---|
Version: | unspecified → git-master |
What makes you believe that user2.mka actually contains a talk at all (all I get is something like "one", followed by "ok" a bit later; apart from that, it is just noise)? Or asked differently, what did you do with MKVToolNix to get proper sound with user2.mka?
(And I can assure you that MKVToolNix does not produce a proper mp3 from this input, as MKVToolNix does not contain an opus decoder and an mp3 encoder.)
PS: These files are btw strange: They lack the audio sample rate and channel count fields in the header.