Drop an episode and get timestamps, chapter titles and show notes
Intermediate16h1–3
WhisperEmbeddingsLLM
16/20Live Talk Captions and Translation
Show live captions of a talk and let each viewer pick their language.
Advanced30h2–4 people16/20 fit
Build Live Talk Captions and Translation for conference and classroom audiences in ~30h with 2–4 people. Stack: whisper.cpp + Socket.IO + LLM + Next.js.
Why it wins: Accessibility plus a real-time wow moment.
Build flow
01 · Problem
Talks are inaccessible to deaf attendees and non-native speakers without expensive captioners.
02 · Solution
Stream microphone audio to whisper.cpp on a laptop, broadcast captions over WebSockets and translate per viewer on their own device.
03 · Build
- 1Run whisper.cpp streaming on a laptop
- 2Broadcast segments with Socket.IO
- 3Translate on the viewer side
- 4Add a QR join page
04 · Demo
Caption the judge's own question live and switch your phone to Hindi.
MVP checklist
- Live captions
- Per-viewer translation
- QR join link
- Transcript download
Tech stack
| Tech | Role | |
|---|---|---|
| whisper.cpp | On-device speech-to-text | AI |
| Socket.IO | Realtime multiplayer | Backend |
| LLM | Language model | AI |
| Next.js | Frontend + API routes | Web |
Reference repos
Open-source projects you can build on. Links checked 2026-10-31.
Related ideas
More AISee where in a call the customer got frustrated and why
Intermediate16h1–3WhisperLLMCharts16/20Practise a spoken interview and get feedback on answers, pace and filler words
Intermediate18h1–3WhisperLLMTTS17/20