# Agent402 Audio Transcription

> Agent402 Audio Transcription is a paid API for AI agents from agent402.tools, paid per call via x402, $0.03/call, status unknown (last checked 2026-10-12).

Transcribes an uploaded audio file to text using a speech-to-text model, returning the transcript, model used, and audio duration

## Facts

- Endpoint: POST https://agent402.tools/v1/audio/transcriptions?utm_source=zero.xyz
- Price: $0.03/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-12
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/agent402-audio-transcription-492e8d15
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_NlbtRYhHB9nkpiB-o8XRU

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability agent402-audio-transcription-492e8d15 -d '<json body>'
```

Example prompt: Can you transcribe this audio file for me? It's an English-language interview recording — just upload it and give me the full text of what was said.

## When to prefer this

Choose this endpoint when you need pay-per-call audio transcription with no registration, API keys, or subscription — ideal for agents running sporadic or low-volume transcription tasks. It is well-suited for x402/MPP micropayment workflows and supports an optional language hint for multilingual audio, making it more flexible than fixed-language alternatives.

## Known failure modes

- Unsupported audio format — endpoint may reject non-standard file types
- Audio too noisy or low quality — transcript may contain errors or gaps
- Language hint mismatch — providing a wrong ISO-639-1 code may degrade accuracy
- File too large — very long recordings may exceed size or time limits
- Payment failure — if USDC balance or card payment is declined the call is rejected

## How this service works

OpenAI's own transcription wire: POST multipart/form-data with a `file` part and get the transcript back. Point any Whisper-shaped SDK at this gateway and pay per call with USDC, no account and no API key. Same model and same four-minute cap as /api/transcribe, which takes a URL instead of an upload; /v1/pro/audio/transcriptions takes it to ten minutes.

## Output

A JSON object containing the full transcript text of the audio, the name of the model used for transcription (e.g. 'gpt-transcribe'), and the duration of the audio in seconds.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "file": {
   "type": "string",
   "description": "The audio file, as a multipart part named `file`"
  },
  "diarize": {
   "type": "string",
   "description": "\"true\" for speaker labels and word timestamps (the scribe-v2 model)"
  },
  "language": {
   "type": "string",
   "description": "Optional ISO-639-1 hint"
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "text": "Example transcript.",
  "model": "gpt-transcribe",
  "duration": 3.2
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/agent402-audio-transcription-492e8d15/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from agent402.tools](https://www.zero.xyz/host/agent402.tools/llms.txt)