It would be great if we can upload an audio file along with a text piece that is the transcription of the audio file (an output from an existing text-to-speech tool). So we use Prodigy’s Audio transcribe recipe for quality assurance and editing rather than transcribing the audio from scratch.
This is already possible!
Assuming you have an examples.jsonl file like below, which has transcriptions from a model filled in beforehand.
{"transcript": "my name is vincent", "audio": "voice.mp4"}
Then you could get what you want by running:
python -m prodigy audio.transcribe xxx examples.jsonl --loader jsonl --fetch-media
Here's what it looks like on my side.
There are a bunch of other settings that may be relevant too, figured I might share the recipe link as well.
Let me know if this does not cover your use-case.
1 Like
This indeed fulfills my needs. Thanks!
1 Like
Happy to hear my suggestion helper. That said ... this support ticket also prompted me to make a new plugin. Figured I'd share ... introducing ... Prodigy-Whisper!
You can explore the plugin repo here:
Let me know if this is helpful or if there's any feedback 
!