ccallazans/ai-video-generator — reverse-engineered prompt

Build me an AI video story generator that I can run locally with Docker. I want to send it a simple prompt, like “tell me a story about Bahia”, and have it automatically write a short story, turn that story into spoken audio, create captions from the audio, then merge the captions onto a background video and save the finished video in a generated folder.

Please make it easy to start with one command, and expose a simple API endpoint where I can POST the message as JSON. Use Ollama locally for the story generation, and include whatever worker or Python pieces are needed for text to speech, captions, and video processing. It should include a few background videos in a resources folder, and pick one randomly for each generated video.

Set it up so the first run can download the needed model and dependencies, but later runs are cached. Keep the code organized and include clear README instructions for Docker and local development.