400-riverside-node

Riverside.fm Recording Transcription

Receive Riverside webhook events when a recording completes, download each participant's isolated audio track, transcribe every track with Deepgram pre-recorded STT, and produce a merged speaker-labelled transcript.

What you'll build

A Node.js Express server that listens for Riverside recording.completed webhooks, fetches the per-track audio files (each participant recorded separately at source quality), submits each track to Deepgram STT with diarization disabled (since tracks are already per-speaker), and returns a merged, time-ordered transcript with speaker labels.

Prerequisites

• Node.js 18 or later
• Deepgram account — get a free API key
• Riverside account — sign up

Environment variables

Copy .env.example to .env and fill in your keys:

Variable	Where to find it
DEEPGRAM_API_KEY	Deepgram console → API Keys
RIVERSIDE_API_KEY	Riverside dashboard → Settings → API

Install and run

cp .env.example .env
# Fill in your API keys in .env

npm install
npm start

The server starts on port 3000 (override with PORT env var).

Endpoints

Method	Path	Description
GET	/health	Health check
POST	/webhook/riverside	Receives Riverside webhook events
POST	/transcribe	Manual transcription — accepts { tracks: [{ participant_name, download_url }] }

Key parameters

Parameter	Value	Description
model	nova-3	Deepgram's most accurate speech model
smart_format	true	Adds punctuation, casing, and paragraph formatting
diarize	false	Disabled — Riverside already isolates speakers into separate tracks
tag	deepgram-examples	Tags usage in the Deepgram console for tracking

How it works

1. Riverside sends a recording.completed webhook with metadata about each participant's audio track
2. The server downloads each track's audio via the Riverside download URL
3. Each track is submitted to Deepgram pre-recorded STT independently (no diarization needed since each track is one speaker)
4. Word-level results from all tracks are merged and sorted chronologically
5. Consecutive words from the same speaker are grouped into segments
6. The final output is a speaker-labelled transcript with timing information

Example output

{
  "transcript": "[Host] Welcome to the show today.\n[Guest] Thanks for having me!",
  "segments": [
    { "speaker": "Host", "start": 0.0, "end": 1.5, "text": "Welcome to the show today." },
    { "speaker": "Guest", "start": 1.6, "end": 2.8, "text": "Thanks for having me!" }
  ],
  "word_count": 11,
  "track_count": 2,
  "speakers": ["Host", "Guest"]
}

Starter templates

deepgram-starters