TeleStudio
Built at Zero to Agent: Vercel x Deepmind Hackathon SF · Mar 21, 2026 · San Francisco, CA

A Next.js application that autonomously generates fully composed, production-ready videos from a single text prompt. It orchestrates a sophisticated, multi-model AI pipeline: Google Gemini 2.5 Flash acts as the "Director" by writing a structured, scene-by-scene script complete with visual descriptions, camera directions, and narration text. Google Veo 3.1 acts as the "Cinematographer", generating high-quality video clips for each scene. ElevenLabs acts as the "Voice Actor", generating realistic TTS narration. Finally, Remotion programmatically stitches these AI-generated assets together—along with text overlays and background music—into a final, exportable MP4. To make the tool incredibly accessible, we integrated a Chat SDK and a Telegram Bot. Users can simply text their video ideas directly to our bot in Telegram. The system handles the entire complex orchestration behind the scenes, processing the prompt, rendering the video on Vercel and Supabase, and delivering the final MP4 back to the user right in the chat. The problem it solves: Traditional video production is a fragmented, expensive, and time-intensive process requiring specialized skills in scriptwriting, stock footage sourcing, voice recording, and timeline editing. This project democratizes video creation by compressing a multi-day studio workflow into a single chat message. It empowers creators, educators, and marketers to generate cohesive, high-quality video content directly from Telegram in minutes without needing any video editing software or production experience.