Error Reference
When a Helios model run completes with status COMPLETE_ERROR, the
response includes two fields to help diagnose the issue:
Error codes
ERR_RECORDING_INVALID_FORMAT
The recording supplied was not in a supported file format.
Supported formats: FLAC, MP3, MP4, Ogg, WebM, WAV
Recommended action: Verify the file format and convert if necessary before submitting a new model run.
ERR_RECORDING_TOO_SHORT
The recording supplied contained less than the minimum required amount of speech.
Minimum speech duration: 10 seconds
Recommended action: Ask the user to speak for longer and submit a new model run.
ERR_RECORDING_TOO_LONG
The recording supplied exceeded the maximum allowed length.
Maximum recording length: 3 minutes
Recommended action: Trim the recording to under 3 minutes and submit a new model run.
ERR_TRANSCRIPTION_FAILED
No speech was detected in the recording supplied.
Recommended action: Check audio quality and ask the user to check their microphone before resubmitting.
ERR_INTERNAL
An internal error occurred during model processing.
Recommended action: Retry the model run. If the error persists, contact support@thymia.ai with the model run ID.
Warnings
Results returned alongside warnings should be treated with caution. Consider flagging these to your users or discarding them depending on your use case.
HTTP errors
A 404 response when retrieving results may indicate that data
deletion was enabled for the model run and the 12-hour retention
window has passed.
Next steps
- Integration Guide — Build a complete production integration including error handling.
- Interpreting Results —
Understand what
COMPLETE_OKresults mean and how to use them. - API Reference — Complete endpoint documentation for the Helios API.






