bifrost-patch-clip-diffusion

Connects multimodal language models with diffusion models using patch-level CLIP embeddings as shared latent variables, enabling controllable image generation with minimal training overhead.

Independent GitHub snapshot · Automated classification · Original documentation governs