Nodes/Comfy EverAnimate/Comfy EverAnimate
ComfyUI Node

Comfy EverAnimate

Native WanAnimate conditioning node with EverAnimate anchor and latent motion memory.

By younestft·Created 2 months ago·Updated 2 months ago· 17
Comfy EverAnimate
  • positive
  • negative
  • vae
  • clip_vision_output
  • reference_image
  • face_video
  • pose_video
  • background_video
  • character_mask
  • continue_motion
  • prev_samples
  • video_anchor_latent
  • positive
  • negative
  • latent
  • trim_latent
  • trim_image
  • video_frame_offset
width832
height480
length77
batch_size1
num_video_anchor_latents4
num_motion_latents1
video_frame_offset0
pose_strength1.000
face_strength1.000
motion_handoff_strength1.00
continue_motion_max_frames5
Categorymodel/conditioning/video_models

Inputs (23)

NameTypeDefaultDescription
positiveCONDITIONING
negativeCONDITIONING
vaeVAE
widthINT83216–16384
heightINT48016–16384
lengthINT771–16384
batch_sizeINT11–4096
num_video_anchor_latentsINT41–16EverAnimate N. The paper/LoRA default is 4 identity anchor latent slots.
num_motion_latentsINT10–16EverAnimate M. Number of previous sampler latents to carry into the next chunk.
video_frame_offsetINT00–16384Connect this from the previous Comfy EverAnimate video_frame_offset output.
pose_strengthFLOAT1.0000–10Multiplier for native WanAnimate pose latents. 0 disables pose influence.
face_strengthFLOAT1.0000–10Strength for native WanAnimate face guide. 0 sends a neutral face guide.
motion_handoff_strengthFLOAT1.000–1How strongly the previous chunk's motion latents are locked into the next chunk. 1.0 is the original hard handoff; 0.6-0.8 is usually smoother for low-step distill runs.
continue_motion_max_framesINT51–16384When continue_motion images are connected, carry this many final RGB frames into the next chunk. This matches native WanAnimate's image carry-over behavior.
clip_vision_outputoptCLIP_VISION_OUTPUT
reference_imageoptIMAGE
face_videooptIMAGE
pose_videooptIMAGE
background_videooptIMAGE
character_maskoptMASK
continue_motionoptIMAGENative-style image carry-over. Connect the previous chunk's final decoded images here. When connected, this takes priority over prev_samples.
prev_samplesoptLATENTPrevious native Wan sampler output. The node takes the last M latents as motion memory.
video_anchor_latentoptLATENTAdvanced: prebuilt N anchor latents. Leave empty to repeat the reference image latent.

Outputs (6)

NameTypeDescription
positiveCONDITIONING
negativeCONDITIONING
latentLATENT
trim_latentINT
trim_imageINT
video_frame_offsetINT