11 upvotes
Video Model Priors Power Unified Image Generation
What is FrameForge: Harnessing Priors of Video Foundation Models for Unified Image Generation?
FrameForge leverages pretrained video foundation models to keep visual cues from an input image while letting a text prompt reshape the output. By treating the image as a single-frame video, it preserves structure and style better than text-only generators.
Who it is for
Researchers and developers building multimodal generation tools can use it to improve consistency between input references and textual edits.