Dev builds Gemini real-time voice demo in 20 minutes, open-sources the repo
AI_Andrew · x · 2026-09-20
Developer AIAndrew built a quick prototype, "Capture the Moment," using Google Gemini's real-time voice stack (Live and Transcribe) — reportedly in about 20 minutes. It's a WebGPU simulation of colored-pencil autumn leaves tumbling onto paper, with live instant transcription and advanced reasoning layered in.
The repo (AndrewVoirol/capture-the-moment) is open source: requires Node.js 18+ and a WebGPU-capable browser (Chrome 113+, Safari 18+); clone, npm install, and run locally. The author sees big potential in real-time voice apps and invites others to fork and play with it.
Related event: Developer Builds Open-Source Gemini Live Voice Demo in 20 Minutes(2 posts)→
More from Apps
- Ethan Mollick prompts Fable with one sentence to build a surprising 'zoom out' game — it works — emollick · 2026-09-20
- OpenAI launches Astra for Law: GPT-6-powered legal AI with daily-updated case law search — minchoi · 2026-09-20
- Veloop: a Chrome extension built with Claude Code that loops your webcam so you can skip video calls — Moist_Tonight_3997 · 2026-09-20
- UX designer with no coding background ships a free iOS chord app built entirely with Claude — muna0001 · 2026-09-20
- Appshots tip: click this to see exactly what the model sees — nickbaumann_ · 2026-09-20
- Redditor Launches a Character-Consistency Prompt Generator for AI Image Creation — Nefilim_666 · 2026-09-20