Unverified50% confidenceFactExact time
Animate Anyone (Alibaba, 2023) was the first to systematically introduce ReferenceNet, injecting reference image appearance features into each frame's generation process via cross-frame attention mechanisms.
1
Sources
50%
Confidence
Long-term
Relevance
7/24/2026
First Seen
Sources
DiffusionGemma角色动作迁移实战：ComfyUI工作流拆解
redditr/comfyui7/12/2026
Related Claims
UnverifiedRepresentative academic works in reference-driven video generation include Animate Anyone, Champ, and MimicMotion, which achieve the separation of identity and motion by introducing dedicated reference encoders (ReferenceNet) or appearance attention mechanisms.74% similarUnverifiedAnimateDiff (2023) achieves controllable animation by inserting temporal attention modules into pre-trained image diffusion models.73% similarVerifiedIP-Adapter (Image Prompt Adapter) was proposed by Tencent Research in 2023; it preserves character identity features by encoding the visual features of a reference image and injecting them into the cross-attention layers of a diffusion model.71% similarUnverifiedThe Doubao large model can be used to reverse-engineer reference images and automatically generate detailed AI drawing prompts.69% similarUnverifiedIP-Adapter allows users to directly input a reference image as a visual prompt to quickly reproduce a character's appearance without training a LoRA.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/607786API
curl https://kongchang.com/api/v1/knowledge/claims/607786MCP
get_claim(id=607786)