Google DeepMind showcases Gemini Audio with live sushi translation demos

Google DeepMind hosted an offline launch event after releasing the Gemini Audio model family this week, and multiple attendees shared enthusiastic hands-on impressions.

Confirmed

• The most popular demo was a sushi stall: participants including jacalulu, petergyang and Google developer relations staffer DynamicWebPaige ordered in English from a Japanese-only vendor, with Gemini Live providing real-time bidirectional translation — the sushi had been flown in fresh from Japan that morning.
• Gemini coached golf swings live; per jacalulu and petergyang, the second swing did improve.
• Lyria composed a song about the evening on the spot and pressed it onto vinyl.
• The event also showed apps built with Gemini TTS, where simple prompts control tone, pacing, emotion and even multiple speakers; jocarrasqueira felt real-time translation use cases like the sushi stall suit consumer-facing scenarios.

Why it matters

These demos vividly demonstrate Gemini audio models' practical capabilities in real-time speech translation, controllable speech generation and interactive coaching, with cross-language real-time communication showing especially strong mass-market potential.

2026-09-25 ~ 2026-09-26 · 5 related posts

• [source] Gemini Live translates English-to-Japanese sushi order in real time at Google's audio event — jacalulu · 2026-09-25
• [source] Gemini Live demoed at Google event, translating English-Japanese sushi order in real time — DynamicWebPaige · 2026-09-25
• [source] Gemini Audio Launch Party: Devs Build Steerable TTS Apps, Sushi Translation Demo Steals Show — jocarrasqueira · 2026-09-25
• Gemini Audio Live demos: bilingual sushi ordering, golf coaching, custom vinyl — jacalulu · 2026-09-26
• Gemini Audio Live demo: ordering sushi in two languages, golf coaching, Lyria pressing a vinyl — petergyang · 2026-09-26