SkyNotSilent /insightcut-jianying-image-video

About insightcut-jianying-image-video

InsightCut turns a one-line topic or a full script into a narrated image video, one image, voiceover and subtitle per segment. You get an MP4, an asset pack, or a draft for Jianying (CapCut's Chinese desktop editor) or CapCut. You can redo one image without starting over. Images come only from the paid Agnes API, and you also need text and voice API keys.

Best for: Creators making narrated explainer or story videos from stills who want to finish the edit in Jianying or CapCut.

• Topic mode writes the full script from one line; script mode keeps your text
• Shows the script, segments, prompts and voice before you spend image credits
• Retries only the images or voiceovers that failed
• Keeps every image and audio version so you can restore an old one
• Exports MP4, a per-segment asset pack, or a Jianying/CapCut draft for Mac or Windows
• Voices with Doubao TTS or Xiaomi MiMo TTS, plus local MiMo voice clones

1. 1
   Install Python 3.11, Node.js 22 and npm, and system FFmpeg and ffprobe. Then the command below and the next command.

   git clone https://github.com/SkyNotSilent/insightcut-jianying-image-video.git

   Copy

   cd insightcut-jianying-image-video

   Copy
2. 2
   Backend: cd ai-kepu-video-server, python3.11 -m venv venv311, source venv311/bin/activate, the command below, cp .env.example .env.

   pip install -r requirements.txt

   Copy
3. 3
   Frontend: the command below and npm install.

   cd ../ai-kepu-video-web/frontend

   Copy
4. 4
   Start the backend in ai-kepu-video-server/ with the command below, and the frontend in another terminal with npm run dev.

   python -m uvicorn api_server:app --host 127.0.0.1 --port 2002 --reload

   Copy
5. 5
   Open http://localhost:2001/settings and add your LiteLLM text model, Agnes image API key and Doubao or MiMo TTS keys.