# Agent Tools Audio Transcription Upload

> Agent Tools Audio Transcription Upload is a paid API for AI agents from agents.crowdspark.me, paid per call via x402, $0.02/call, status unknown (last checked 2026-11-01).

Transcribes or translates uploaded audio files (MP3, WAV, MP4, OGG, FLAC, WebM) to text with language detection, timestamps, segments, and WebVTT output

## Facts

- Endpoint: POST https://agents.crowdspark.me/v1/audio/transcribe/upload?utm_source=zero.xyz
- Price: $0.02/call
- Payment: x402
- Status: unknown
- Last checked: 2026-11-01
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/agent-tools-audio-transcription-upload-d753f47f
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_9M2gS8jdjD33s8z0kL6TX

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 agent-tools-audio-transcription-upload-d753f47f -d '<json body>'
```

Example prompt: Can you transcribe this MP3 audio file I'm uploading? Use the 'transcribe' task and detect the language automatically — I need the full text, timestamps for each segment, and WebVTT captions.

## When to prefer this

Choose this endpoint when you have raw audio bytes available locally or in memory and need a full-featured transcription response including language detection, timestamped segments, and WebVTT captions in a single call. It supports multiple upload methods (raw body, multipart form, or base64 JSON), making it flexible for agent pipelines. Prefer it over URL-based transcription endpoints when you already have the file on hand rather than a remote URL, and when you need structured segment data alongside the plain transcript.

## Known failure modes

- File exceeds 10 MB limit — request rejected with size error
- Unsupported audio format — only audio/mpeg, audio/wav, audio/mp4, audio/ogg, audio/flac, audio/webm accepted
- Audio too short or silent — may return empty text or minimal segments
- Invalid language ISO code — may fall back to auto-detection or return an error
- Payment failure or insufficient USDC — x402 payment required before processing
- Corrupted or malformed audio bytes — decoding error returned
- Network timeout on large files near the size limit

## How this service works

Pay-per-call utilities for AI agents over x402, USDC on Base.

## Output

A JSON object containing the task type, full transcribed or translated text, audio byte size, model used (whisper-large-v3-turbo), audio format, detected language code, an array of segments each with start time, end time, and text, total word count, duration in seconds, and WebVTT-formatted caption data.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "task": {
   "enum": [
    "transcribe",
    "translate"
   ],
   "type": "string"
  },
  "language": {
   "type": "string",
   "description": "Optional ISO code, e.g. en"
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "task": "transcribe",
  "text": "Hello and welcome to the show.",
  "bytes": 67000,
  "model": "whisper-large-v3-turbo",
  "format": "mp3",
  "language": "en",
  "segments": [
   {
    "end": 4.2,
    "text": "Hello and welcome to the show.",
    "start": 0
   }
  ],
  "word_count": 6,
  "duration_seconds": 4.2
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/agent-tools-audio-transcription-upload-d753f47f/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from agents.crowdspark.me](https://www.zero.xyz/host/agents.crowdspark.me/llms.txt)