raojiacui /prompt-lens

About prompt-lens

Upload a clip and PromptLens hands back a prompt you can reuse. A vision model reads the scene, style, characters and motion. You can transcribe audio and cut clips too, but cutting needs your own FFmpeg server. Docs are Chinese-only, and you'll need paid API keys plus a PostgreSQL database.

Best for: Prompt researchers studying reference clips before they shoot their own version.

• Analyzes an uploaded video and writes a prompt for scene, style and action
• Handles a single image or a batch of images
• Works with Zhipu AI, Google Gemini, OpenRouter and Kie.ai keys
• Transcribes dialogue and splits it into segments with an LLM
• Exports trimmed or joined clips through your own FFmpeg service
• Keeps prompt history and separates each user's data

1. 1
   Clone the repo: the command below then cd prompt-lens

   git clone https://github.com/raojiacui/prompt-lens.git

   Copy
2. 2
   Install dependencies with pnpm install
3. 3
   Copy the env template: cp .env.example .env.local, then fill in your database, R2 and AI key settings (see LOCAL_SETUP.md)
4. 4
   Set up the database with pnpm db:push
5. 5
   Start the dev server with pnpm dev and open http://localhost:3000
6. 6
   Log in, open Settings, pick a provider (Zhipu AI / Gemini / OpenRouter / Kie.ai) and save your key