Introduction
Transcribe Copilot is a self-hosted, private AI meeting transcription and intelligence platform built to replace expensive SaaS subscriptions (like TurboScribe at 20/year) with direct developer APIs and zero-scale serverless containers.
You can run and use the live application at https://transcribe.aldianapps.com.
Key Features
- Multimodal Video & Visual OCR: Continuously decodes video streams at 1 FPS with Gemini 3.7 Flash, detects slide/screen transitions, extracts on-screen code/diagram OCR, and captures synchronized presentation slides in the browser using HTML5 Canvas.
- Multi-Engine Routing: Support for Google Gemini 3.7 Flash, Gemini 2.5 Flash, Groq Whisper LPUs (1 hour audio transcribed in ~10 seconds), OpenAI Whisper, OpenRouter, and self-hosted endpoints.
- Client-Side BYOK Key Vault: API keys are encrypted client-side using Web Crypto AES-GCM (256-bit) + PBKDF2 salt derivation. Keys and raw audio stay entirely within the user's browser.
- Meeting Intelligence: Generates executive summaries, key discussion points, checkable action items with owners/deadlines, and decision logs.
- Rich Multi-Format Exports: Word (.docx), Markdown (.md), Subtitles (.srt, .vtt), Plain Text, and JSON.
Architecture & Hosting
- Frontend & App Framework: Next.js 16 (App Router), React 19, TypeScript, and Tailwind CSS v4 compiled to standalone mode.
- Infrastructure & Serverless Hosting: Google Cloud Run with scale-to-zero autoscaling (0 CPU/RAM when idle, bash/month base cost) in region
us-central1. - CI/CD Pipeline: Automated Google Cloud Build triggers deploying Docker images through Google Artifact Registry with custom domain mapping at
transcribe.aldianapps.com.