feat(ai): apply code improvements to AudioToText pipeline #4
Conversation
| @@ -75,6 +75,9 @@ var ( | |||
| ErrProfEncoder = fmt.Errorf("unknown VideoProfile encoder for protobufs") | |||
| ErrProfName = fmt.Errorf("unknown VideoProfile profile name") | |||
|
|
|||
| ErrUnsupportedAudioFormat = fmt.Errorf("audio format unsupported") | |||
There was a problem hiding this comment.
@eliteprox this is a very usefull error which should be implemented but currently is not used. I tried inputting the following unsupported audio OGG filetype:
But it threw a audio duration calculation error. Can we implement a audio file type unsupported error?
There was a problem hiding this comment.
Checking file extensions isn't super reliable and ffmpeg has some edge cases where it returns 0 duration for supported file types, so we currently just check if the duration is <= 0.
There was a problem hiding this comment.
The status code returned from GetCodecInfoBytes (which currently ignored) should indicate something that indicates the file format is unsupported, although I don't know exactly what the code would be. However I am unsure why the API returns both a code and an error; ideally it should be one or the other.
For what it is worth in this specific case, ffmpeg should be able to detect the duration for this file. Ogg is probably just not compiled into in the production LPMS build, which has a slimmed-down version of ffmpeg. Adding the ogg demuxer to install_ffmpeg.sh might be enough for this case.
rickstaa
force-pushed
the
add-speech-to-text-code-improvements
branch
from
3e88726
to
5104725
Compare
This commit applies several code improvements to the AudioToText codebase.
rickstaa
force-pushed
the
add-speech-to-text-code-improvements
branch
from
5104725
to
cb4360b
Compare
@eliteprox here are some small code improvements I found.