The catalog is public and key-free so your code can discover models at runtime:
GET https://flopco.in/v1/models
Filter with ?type=text|image|video|audio and ?tag=vision|edit|pixel|face-swap|talking-head|character-swap|.... Pick the encoding with ?format=json (default), xml, or php.
Each model reports id, type, family (talking-head, reenact, face-swap, character-swap, generator, pixel, edit, vector), capabilities (e.g. image-understanding for vision), acceptsInitImage, cost (in FLOP), license, commercialOk (false = research-only weights), and a status of live (served now) or preview (listed, wiring pending).
Live text + vision models today include GPT-4o plus self-hosted Qwen2.5-VL, LLaVA, Llama 3.2 Vision, and Gemma 3 - all callable for image-to-text.