M4A conversion fails in audio-to-text pipeline

[thomshutt]

Describe the bug

When passing an m4a file into the audio-to-text pipeline, it errors with:

2024-10-04 12:39:05,158 - app.routes.audio_to_text - ERROR - Audio processing error: Could not determine output format

Passing the same (converted) file in as an MP3 succeeds.

Reproduction steps

Run docker run -d --restart unless-stopped --name runner-audio-to-text-test -v /home/user/models:/models:rw -e PIPELINE=audio-to-text -e MODEL_ID=openai/whisper-large-v3 -p 9000:9000 livepeer/ai-runner:latest uvicorn app.main:app --log-config app/cfg/uvicorn_logging_config.json --host 0.0.0.0 --port 9000

Call curl -X POST http://0.0.0.0:9000/audio-to-text -F model_id="openai/whisper-large-v3" -F [email protected]

Expected behaviour

Expect it to work and produce text

Severity

Minor

Screenshots / Live demo link

No response

OS

Linux

Running on

Docker

AI-worker version

latest

Additional context

No response