open-webui/open-webui · error · HTTPException

Audio conversion failed. Chat completions API requires mp3 o

Error message

Audio conversion failed. Chat completions API requires mp3 or wav format.

What it means

Raised (500) when Mistral STT runs in chat-completions mode (`audio.stt.mistral.use_chat_completions` true) and the helper `convert_audio_to_mp3(file_path)` returns a falsy value. That helper shells out through pydub/ffmpeg; it returns None when ffmpeg is missing, the binary crashes, or the input audio is unreadable. Since the chat/completions payload embeds base64 audio that Mistral only accepts as mp3 or wav, the handler aborts when conversion fails.

Source

Thrown at backend/open_webui/routers/audio.py:957

    r = None
    try:
        model = await Config.get('audio.stt.model') or 'voxtral-mini-latest'
        log.info(
            f'Mistral STT - model: {model}, method: {"chat_completions" if use_chat_completions else "transcriptions"}'
        )

        session = await get_session()
        if use_chat_completions:
            audio_file_to_use = file_path
            if is_audio_conversion_required(file_path):
                log.debug('Converting audio to mp3 for chat completions API')
                converted_path = await asyncio.to_thread(convert_audio_to_mp3, file_path)
                if converted_path:
                    audio_file_to_use = converted_path
                else:
                    log.error('Audio conversion failed')
                    raise HTTPException(
                        status_code=500,
                        detail='Audio conversion failed. Chat completions API requires mp3 or wav format.',
                    )

            async with aiofiles.open(audio_file_to_use, 'rb') as audio_file:
                raw = await audio_file.read()
                audio_base64 = {
                    'data': base64.b64encode(raw).decode('utf-8'),
                    'format': mimetypes.guess_extension(mimetypes.guess_type(audio_file_to_use)[0]).lstrip('.'),
                }

            language = metadata.get('language', None) if metadata else None
            text_instruction = (
                f'Transcribe this audio exactly as spoken in {language}. Do not translate it.'
                if language
                else 'Transcribe this audio exactly as spoken in its original language. Do not translate it to another language.'
            )

View on GitHub (pinned to 01f4282f1f)

Solutions

1. Install ffmpeg on the server running Open WebUI (apt-get install -y ffmpeg) and confirm `ffmpeg -version` works as the app user.
2. If ffmpeg cannot be installed, disable chat-completions mode: set audio.stt.mistral.use_chat_completions to false so the multipart /audio/transcriptions endpoint is used, which accepts other formats.
3. Verify the uploaded file is playable/complete (ffprobe file.webm) and re-upload; a 0-byte file means the upload itself failed.
4. Check server logs for the preceding 'Audio conversion failed' log.error line to confirm which input path failed.

Example fix

# before: engine=mistral + use_chat_completions=true, host without ffmpeg
# -> 500 'Audio conversion failed. Chat completions API requires mp3 or wav format.'

# after (option A): install decoder
# apt-get update && apt-get install -y ffmpeg

# after (option B): switch to transcriptions mode
await Config.save('audio.stt.mistral.use_chat_completions', False)

import shutil, mimetypes

def can_use_chat_completions_stt(path: str) -> bool:
    mime = mimetypes.guess_type(path)[0]
    needs_conversion = mime not in ('audio/mpeg', 'audio/wav', 'audio/x-wav')
    if needs_conversion and not shutil.which('ffmpeg'):
        return False  # conversion will fail -> avoid 500
    return True

from fastapi import HTTPException
try:
    result = await transcribe(request, path, metadata, user)
except HTTPException as e:
    if 'conversion failed' in str(e.detail).lower():
        if not shutil.which('ffmpeg'):
            raise RuntimeError('ffmpeg missing: disable use_chat_completions or install ffmpeg') from e
    raise

When it happens

Trigger: Uploading browser-recorded audio (audio/webm from MediaRecorder) or .ogg/.m4a with use_chat_completions enabled, on a host where ffmpeg is not installed or not on PATH; or uploading a zero-byte/corrupt file that pydub cannot decode.

Common situations: Docker images or slim Debian containers without the ffmpeg package; ffmpeg present but an incompatible/old version; a truncated upload (proxy cut the body) producing a non-decodable file; works locally (ffmpeg installed) but fails in CI or production container.

Related errors

• Audio conversion failed. The audio file may be corrupted or
• Failed to parse Mistral response: {str(e)}
• [ERROR: Error processing audio file]
• Mistral API key is required for Mistral STT
• Error transcribing chunk: {transcribe_exc}

AI-assisted analysis of open-webui/open-webui@01f4282f1f (2026-08-14). Data as JSON: /api/errors/496bdeeef28da966. Report an issue: GitHub.