Affected page or section
https://github.com/deepgram/deepgram-python-sdk#file-transcription
What is unclear or wrong?
The File Transcription example (and its async/error-handling/raw-response variants) calls transcribe_file with only model="nova-3", no language. If the audio isn't in English, this doesn't raise an error — it silently returns a near-empty, low-confidence transcript (e.g. a single hallucinated word at confidence 0.33 for an 11-second non-English clip). This is easy to miss since there's no error to signal the problem.
Suggested change
Add a note in the File Transcription section stating that nova-3 assumes English by default when language isn't set, and show how to set it explicitly (e.g. language="fr", or language="multi" for auto-detection).
Example code (if any)
Colle un extrait minimal, par ex. : response = client.listen.v1.media.transcribe_file(request=audio_file.read(), model="nova-3", language="fr")
SDK parity (if relevant)
Session ID (optional)
No response
Project ID (optional)
No response
Request ID (optional)
No response
Affected page or section
https://github.com/deepgram/deepgram-python-sdk#file-transcription
What is unclear or wrong?
The File Transcription example (and its async/error-handling/raw-response variants) calls transcribe_file with only model="nova-3", no language. If the audio isn't in English, this doesn't raise an error — it silently returns a near-empty, low-confidence transcript (e.g. a single hallucinated word at confidence 0.33 for an 11-second non-English clip). This is easy to miss since there's no error to signal the problem.
Suggested change
Example code (if any)
SDK parity (if relevant)
Session ID (optional)
No response
Project ID (optional)
No response
Request ID (optional)
No response