x402_transcribe_audio
Transcribe audio files from URLs into text with automatic language detection, supporting MP3, WAV, M4A, FLAC, OGG and other formats. Returns transcript text, detected language, timestamps, and duration.
Instructions
Transcribe an audio file from a URL using faster-whisper with auto language detection. Price: $0.05 USDC per transcription (paid mode) | Free test: returns fixture data.
Supports: MP3, WAV, M4A, FLAC, OGG, and most audio formats. Limits: 25MB file size, 10-minute duration. Payment is charged on download; duration refusals are still charged. Note: transcription can take 30–120 seconds for longer files (CPU-based, requests queue serially). Without X402_PRIVATE_KEY, only the free test endpoint is available.
Returns: transcript text, detected language, language confidence, duration, and segment or word timestamps.
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| url | Yes | URL of the audio file to transcribe (public, http/https, max 25MB, max 10 min) | |
| language | No | ISO 639-1 language hint (e.g. 'en', 'fr', 'es') — omit for auto-detection | |
| word_timestamps | No | Return word-level timestamps instead of segment-level (default: false) |
Implementation Reference
- src/index.ts:696-738 (handler)The tool 'x402_transcribe_audio' is registered and implemented in src/index.ts. It handles both paid (via X402_PRIVATE_KEY) and free test modes.
server.tool( "x402_transcribe_audio", `Transcribe an audio file from a URL using faster-whisper with auto language detection. Price: $0.05 USDC per transcription (paid mode) | Free test: returns fixture data. Supports: MP3, WAV, M4A, FLAC, OGG, and most audio formats. Limits: 25MB file size, 10-minute duration. Payment is charged on download; duration refusals are still charged. Note: transcription can take 30–120 seconds for longer files (CPU-based, requests queue serially). Without X402_PRIVATE_KEY, only the free test endpoint is available. Returns: transcript text, detected language, language confidence, duration, and segment or word timestamps.`, { url: z.string().url().describe("URL of the audio file to transcribe (public, http/https, max 25MB, max 10 min)"), language: z.string().optional() .describe("ISO 639-1 language hint (e.g. 'en', 'fr', 'es') — omit for auto-detection"), word_timestamps: z.boolean().default(false) .describe("Return word-level timestamps instead of segment-level (default: false)"), }, async (params) => { const base = APIS.transcription.baseUrl; try { const usePaid = !!PRIVATE_KEY; if (usePaid) { const payload: Record<string, unknown> = { url: params.url, word_timestamps: params.word_timestamps, }; if (params.language) payload.language = params.language; const data = await apiPost(base, "/transcribe", payload, true); return textResult({ mode: "paid", cost: "$0.05", ...data }); } else { const data = await apiGet(base, "/transcribe/test"); return textResult({ mode: "free_test", note: "Free test — returns fixture data. Set X402_PRIVATE_KEY for real audio transcription.", ...data, }); } } catch (err: any) { return errorResult(err.message); } } );