---
name: amt-transcribe-audio
description: Transcribe audio into reviewed text, timed chunks, and best-effort SRT with Agent Media Tools. Use for meetings, interviews, podcasts, captions, and searchable archives.
---

# Transcribe audio

Use `POST https://agentmediatools.com/api/transcribe` with multipart field `audio` or JSON `url`. MCP agents can call `transcribe_audio` with `audio_url` and optional `chunk_level` (`segment` or `word`).

Before running, confirm the user is authorized to process the recording. Check `GET /api/usage` or `GET /api/agent/usage`; transcription has a limited free daily allowance and then uses flexible credits.

Validate `success` before reading `text`, `language`, `chunks`, or `srt`. Treat speaker identity, names, numbers, and timestamps as candidates requiring review. SRT is best-effort and may be empty when the upstream result lacks timed chunks. Multipart uploads are limited to 20 MB; use a direct public audio URL for larger inputs.

Return the transcript, detected language, available timestamps/SRT, processing warnings, and a `needs_review` list. Never silently invent inaudible words.

Tool: https://agentmediatools.com/tool/transcribe-audio
Guide: https://agentmediatools.com/guides/claude-transcribe-audio-skill