Viggle Remake open weights released for recasting and editing whole videos
Viggle released open weights on Hugging Face for its Remake model, which recasts people or edits existing videos while keeping source motion and soundtrack.
Viggle has released open weights for Remake on Hugging Face, a model that recasts people in video or edits existing footage while keeping the source motion, camera movement and soundtrack. The model is based on Viggle's H3 model, and the release is hosted under the Viggle/Viggle-Remake repository. Viggle also runs a hosted app at viggle.ai.
The release is better understood as a pipeline than as a single model. Its MCP server splits a long video into windows, paints keyframes, generates prompts, renders each segment, checks its own output and stitches the pieces back together. The stitching step turns the segments back into one continuous video. The package also includes ComfyUI nodes and workflows, int8 and GGUF variants for smaller GPUs, and lower-rank LoRAs at rank-128 and rank-64.
One user found Remake strong at replacing generic characters or subjects defined by a prompt, and gave robots as an example. On real people, the same user said Remake is weaker than Minimax H3 at keeping recognizable facial features, and described the output as looking like 'a bad lora version' of the target person. Users report the same pattern in general, but these are informal user reports and have not been independently tested.
A commenter asked whether Remake is a wrapper around an existing video model with added masking and inpainting stages, or whether it uses a distinct generation architecture. The discussion doesn't settle that question, and the release materials as quoted in the thread don't say which it is.
What changed
Viggle released open weights for Remake on Hugging Face, a model that recasts people in video or edits existing footage while keeping source motion and soundtrack.
Type: New capability · Applies to: Remake, Viggle/Viggle-Remake
Why it matters
Open weights let studios and developers run whole-video editing on their own hardware, though identity preservation for real people remains a weak point.
What you can do
Test identity retention on your own footage before relying on the tool for real-person edits.
Original source
We have no link to the original source for this story yet.
Where we saw it
Latent Space
AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works