Recipes

Audio transcript generation

Convert spoken audio into structured, searchable text using Meridian's transcription pipeline.

Overview

This recipe walks through uploading an audio file, selecting a language model, and receiving a timestamped transcript with speaker diarization. Ideal for meeting notes, interviews, and content repurposing.

Prerequisites

  • Active Meridian subscription with transcription quota
  • Audio file in MP3, WAV, or M4A format (max 2 GB)
  • API key with transcribe:write scope

Step-by-step

1. Upload audio

POST to /v1/audio/upload with multipart form data. The response includes a file ID valid for 24 hours.

2. Create transcription job

Send the file ID and desired options to /v1/transcriptions. Enable diarization for multi-speaker audio.

3. Poll for completion

Query /v1/transcriptions/:id until status is completed. Webhook callbacks are also supported.

4. Retrieve transcript

Download the JSON result containing segments, timestamps, and speaker labels. Export to SRT or VTT for subtitles.

Next steps