About this project

LazyEdit provides a comprehensive pipeline for video production, integrating AI-driven generation and post-processing. The workflow encompasses video upload or generation (via integrations like Sora and Veo), ASR-based transcription, multilingual subtitle translation and polishing, subtitle burn-in with layout controls, keyframe captioning, and metadata generation. Technically, the project consists of a Tornado-based API backend and an Expo-powered frontend for web and mobile. It is designed for local-first deployment on Linux, utilizing tmux for session management and optional systemd services for background operation. The system integrates several specialized submodules for ASR (Whisper), image captioning (ViT-GPT2, CLIP-GPT), and an optional AutoPublish module for social media distribution. Key capabilities include: - Prompt-based video generation (Stage A/B/C). - Full media pipeline: transcription -> translation -> burn-in -> captioning -> metadata. - Multilingual support including furigana, IPA, and romaji. - API-first architecture allowing for scripted integration and automation. - Local-first runtime with PostgreSQL support for data persistence.