neopen /story-shot-agent

About story-shot-agent

PenShot reads a script and breaks it into shots with timed video prompts. It writes prompts in Chinese and English for Sora, Veo, Runway and Kling. Memory plus a Chroma vector store keep characters and scenes steady. It's tuned for Chinese scripts, and paid LLM keys plus Redis are required.

Best for: Writers turning long scripts into model-ready shot prompts for Sora, Veo, Runway or Kling.

• Reads film, short drama, anime or novel scripts and splits long text
• Cuts the story into shots and gives each one a duration
• Writes Chinese and English visual prompts, negative prompts and audio prompts
• Keeps characters, scenes and plot steady with tiered memory and Chroma search
• Works with OpenAI, Qwen, DeepSeek, Ollama and other OpenAI-compatible APIs
• Plugs in as a Python SDK, REST API, MCP server, LangGraph node or A2A agent

1. 1
   Install the package: pip install penshot
2. 2
   Copy the env template: cp .env.example .env
3. 3
   Fill in your LLM and embedding keys in .env, plus PENSHOT_REDIS_URL
4. 4
   Run the SDK: the command below, then agent.breakdown_script_async(script)

   from penshot.api import create_penshot_agent

   Copy
5. 5
   Or start the MCP server: the command below

   python -m penshot.mcp_server --max-concurrent 5 --queue-size 500

   Copy
6. 6
   Or deploy with Docker: docker compose up -d