Repository navigation
[Suggestion] WhatsApp Business voice message transcription (Node.js) #189
Description
What to build
A working example that receives WhatsApp voice messages via the WhatsApp Business API (Cloud API), downloads the audio, transcribes it with Deepgram, and replies with the transcript text in the same chat.
Why this matters
WhatsApp is the world's most-used messaging platform with 2B+ users, and voice messages are one of its most popular features. Businesses and developers building WhatsApp bots frequently need to process voice messages — for customer support automation, accessibility, note-taking, and more. A reference example showing the complete flow from WhatsApp webhook to Deepgram transcription to reply message fills a common developer need.
Suggested scope
- Language: TypeScript (Node.js)
- Framework: Express + WhatsApp Cloud API (Meta)
- Deepgram APIs: Pre-recorded STT (Nova-3)
- What it does: Express webhook server that handles incoming WhatsApp messages, detects voice messages, downloads audio via WhatsApp Media API, transcribes with Deepgram, and sends the transcript back as a WhatsApp text reply
- Includes: ngrok setup instructions for local testing, environment variable configuration for WhatsApp Business API credentials
- Complexity: Medium
Acceptance criteria
- Runnable with minimal setup (clone, add API keys, run with ngrok)
- README explains the pattern clearly including WhatsApp Business API setup
- Uses current SDK version
- Handles both audio/ogg and audio/mpeg voice message formats
- Includes error handling for failed downloads or transcriptions
Raised by the DX intelligence system.
Activity
github-actions commentedon Apr 5, 2026
@deepgram/devrel — new suggestion from @deepgram-robot (no write access). Add the
approvedlabel to process it, or close to decline.github-actions commentedon Apr 6, 2026
This looks like it's already covered — see:
That PR is building exactly what you're describing: a Node.js/Express webhook server that receives WhatsApp voice messages, transcribes them with Deepgram, and replies with the transcript.
Feel free to add more context there if you have a different angle in mind!