NarratoAI 🧩 Workflow Open source

What it is

NarratoAI by linyqh (MIT) automates narrated commentary videos. A vision-capable LLM analyses the source footage, writes a narration script, and the tool cuts matching clips, generates the voice-over and subtitles, and renders the final video. It was built for film, documentary and short-drama commentary, a popular format on Chinese and global video platforms, and version 0.8.x arrived in July 2026.

Key features

• Frame analysis with vision models to understand footage
• LLM-written narration scripts (DeepSeek, Gemini, Qwen and OpenAI-compatible APIs)
• Automatic clip selection and editing to match the script
• TTS voice-over including Doubao TTS and other engines
• Subtitle generation and burn-in
• Streamlit web UI, Docker and Windows integration package

How to install

Clone with git clone https://github.com/linyqh/NarratoAI.git, then docker compose up -d, or install dependencies with uv sync and start the UI with uv run streamlit run webui.py --server.maxUploadSize=2048.

Pricing and rights

Free under MIT; you pay for the LLM, vision and TTS APIs you configure. Commentary videos built on films or dramas you do not own raise copyright questions; fair-use rules differ by country and platform. The README also promotes sponsor sign-up codes.

Who it is for

Commentary and recap creators working with footage they own or have licensed, and editors who want a first rough cut from long material.

Verdict

A clever end-to-end narration pipeline that saves hours of rough cutting. Results need editing, documentation is mainly Chinese, and the headline use case carries real copyright risk.