Apr 16
Loading video player...
Thor from Google DeepMind walks through how to build a Telegram bot that receives voice messages, reasons over them with Gemini 3.1 Flash, and responds with generated speech in a custom accent, all built with Antigravity and deployed to Cloud Run without writing the code manually. What's covered: Setting up a Telegram bot with BotFather, using Gemini Flash for reasoning and Gemini TTS for voice responses, handling OGG audio from Telegram and converting output with FFmpeg, building agent, transcription, and translation modes, deploying to Google Cloud Run with Docker and Secret Manager, and using the Gemini Interactions API coding skill to accelerate the build. Grab your Gemini API key at Google AI Studio and your bot token from BotFather on Telegram to get started. Full code and tutorial linked below. Resources: Find the tutorial → https://goo.gle/4cdOpJa Find the code → https://goo.gle/3Q8prCo What are you building with Gemini? Drop it in the comments. Chapters 0:00 - Intro and Telegram voice bot demo 1:17 - Technical Setup and Prerequisites 1:48 - Deploying to Google Cloud Run 2:27 - AI-Assisted Coding with Antigravity 4:08 - Using Gemini 3.1 Flash Lite and TTS 4:51 - Bot Modes: Agent, Transcription, and Translation 6:32 - Voice Message Flow and Audio Conversion 8:31 - GitHub Repo and Next Steps Subscribe to Google for Developers → https://goo.gle/developers Speaker: Thor Schaeff Products Mentioned: Google AI, Gemini