Nodes/ComfyUI-WanAnimatePlus/WanAnimatePlus EverAnimate Embeds
ComfyUI Node

WanAnimatePlus EverAnimate Embeds

EverAnimate conditioning bundle for segmented WanAnimatePlus sampling.

By wuwukaka·Created 3 months ago·Updated 22 days ago· 402
WanAnimatePlus EverAnimate Embeds
  • vae
  • pose_images
  • face_images
  • anchor_images
  • bg_images
  • mask
  • image_embeds
width832
height480
num_frames734
force_offloadtrue
frame_window_size77
pose_strength1.000
face_strength1.000
num_video_anchor_latents4
num_motion_latents1
num_overlap_frame4
use_pingpongtrue
use_image_anchortrue
use_random_frame_anchortrue
random_anchor_with_user_firsttrue
use_repeat_anchorfalse
tiled_vaefalse
CategoryWanAnimatePlus

Inputs (22)

NameTypeDefaultDescription
vaeWANVAE
widthINT83264–8096Width of the video to generate
heightINT48064–8096Height of the video to generate
num_framesINT7341–10000Total output frames. Segment count is computed from this, frame_window_size, and num_overlap_frame.
force_offloadBOOLEANtrueOffload VAE after encoding to save VRAM
frame_window_sizeINT771–10000Effective output frames per EverAnimate segment. Must be 1 mod 4, e.g. 77, 81, 85.
pose_strengthFLOAT1.0000–10Additional multiplier for the pose adapter
face_strengthFLOAT1.0000–10Additional multiplier for the face adapter
pose_imagesIMAGEPose control video
face_imagesIMAGEFace control video. Resized to 512x512 internally.
num_video_anchor_latentsINT41–4Number of anchor latent slots prepended to each EverAnimate segment
num_motion_latentsINT10–4Number of previous-segment motion latents used for continuity
num_overlap_frameINT40–10000Overlapping output frames between adjacent segments
use_pingpongBOOLEANtruePing-pong extend pose, face, bg, and mask sequences when more frames are needed
use_image_anchorBOOLEANtrueUse generated frames from the first segment to build later video anchors
use_random_frame_anchorBOOLEANtrueRandomly sample generated segment-0 frames for later video anchors
random_anchor_with_user_firstBOOLEANtrueIn random-anchor mode, reserve the first manual anchor frame as the user anchor
use_repeat_anchorBOOLEANfalseWhen fewer manual anchor frames are provided than anchor slots, repeat the provided sequence to fill the missing slots
anchor_imagesIMAGEManual identity/anchor image frames. Video inputs are treated as an anchor-frame sequence, not as source-video conditioning.
bg_imagesoptIMAGEOptional background/inpaint video
maskoptMASKOptional mask paired with bg_images
tiled_vaeoptBOOLEANfalseUse tiled VAE encoding for reduced memory use

Outputs (1)

NameTypeDescription
image_embedsWANVIDIMAGE_EMBEDS