Open-weights video model that replaces a character in a clip using one repainted frame, fine-tuned from MiniMax-H3.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Viggle Animate is a video-to-video character replacement model from Viggle. It takes a driving video and one of that video's own frames in which the character has been repainted in an image editor, then propagates the edit across the shot while keeping motion, camera and timing. The video stage uses no pose estimator, segmenter, face tracker or text encoder, and the model is not told what the new character is.
According to the model card, it is a 33.1B full finetune of MiniMax-H3's ref2va transformer, jointly distilled with DMD to three forward passes. The authors report 124 frames rendered in 26 seconds on a B200 GPU, about 6.1x faster per clip than Wan2.2-Animate-14B in their matched comparison. They emphasise fast motion such as head whips, kicks and jumps. Creator coverage mentions full weights of about 66 GB and a 2.5 GB LoRA.
Weights are on Hugging Face under the MiniMax-H3 community license (listed as "other"), with a demo Space and community ComfyUI nodes. The card notes that Viggle has released a newer model, Viggle Remake, which handles multiple people, objects and backgrounds. The model needs a separately prepared edited frame and a large GPU for local use.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Swaps the character in a clip using one edited first frame, without pose estimation or segmentation. It is a fine-tune of Minimax H3 with released full weights (66 GB) and a 2.5 GB LoRA.
Closest neighbours by meaning, not just by category.
Model · Viggle
Geometry-guided video-to-video model from Viggle that re-renders existing footage along a new camera path and timing.
Model · Z.ai
Research framework for end-to-end controlled character animation and motion transfer from a driving video to a reference character.
Model · Meituan
Open-weights audio-driven avatar video model from Meituan, built on LongCat-Video, with 8-step distilled inference.
Model · Alibaba
Alibaba's diffusion-transformer model that animates a character image from a driving video, with a real-time Lite variant.
Model · JD
Open-weights 16B autoregressive diffusion model from JD for real-time, instruction-guided editing of streaming video.
Model · ByteDance
ByteDance's open-weights video generation and editing model pairing a Qwen2.5-VL planner with a Wan2.2 renderer.
