Add your own API

Alpha. these are Hathor’s first videos, made on our own servers or on your engines. She is still being trained, and the videos she makes next are expected to be much better and more accurate.

Every stage of a film can run on your own engine. Pick it in the Engines panel on the create form; your key stays in your browser and is sent only to that provider.

Script & shot plan — any LLM

OpenAI, Gemini, Groq, OpenRouter and Anthropic are built in. For Your own API, give an HTTPS URL that accepts POST {"prompt": "…"} (with your bearer token if you set one) and returns {"text": "…"}, where the text is the JSON plan below. The prompt we send asks for exactly this shape:

{
 "title": "Kush and the Nile",
 "minutes": 10,
 "style": "eerie",
 "chapters": [
  {
   "title": "The river",
   "shots": [
    {
     "image": "https://…/shot1.jpg",
     "card": "Meroë, Kingdom of Kush, c. 300 BC",
     "question": "Who built these?",
     "credit": "Excavation reports, Meroë",
     "secs": 8
    },
    {
     "prompt": "photorealistic recreation of…",
     "card": "",
     "secs": 8
    }
   ]
  }
 ],
 "sources": [
  "…"
 ]
}

A shot needs an image (HTTPS URL or data:image/…;base64) or a prompt (we make the image). secs 2–30. Up to 600 shots.

Images — any image API

fal.ai, Gemini and your own worker come from your Engines settings. For Your own API: HTTPS URL, POST {"prompt": "…", "size": "768x512"} → {"url": "https://…"} or {"base64": "…", "mime": "image/png"}.

Run the whole thing yourself

Our worker is open source. It pulls jobs from https://hathor.soapbox.community/video-studio/api/worker/next with a token, renders with FFmpeg, and sends the film back — see integrations/video_studio_worker.py in the repository.