ComfyUI Node
WanAnimatePlus SCAIL_2 Embeds
A ComfyUI node in WanAnimatePlus with 25 inputs and 1 output.
WanAnimatePlus SCAIL_2 Embeds
- vae
- clip_embeds
- ref_image
- bg_image
- pose_images
- prefix_frames
- prefix_mask
- transition_video
- pose_image_mask
- reference_image_mask
- image_embeds
◄width832►
◄height480►
◄num_frames81►
◄frame_window_size81►
◄force_offloadtrue►
◄pose_strength1.000►
◄ref_strength1.000►
◄replacement_modefalse►
◄tiled_vaefalse►
◄transition_colormatchdisabled►
◄loop_colormatch_referenceprevious_matched_frame►
◄prefix_alpha_cropfalse►
◄preserve_main_ref_backgroundtrue►
◄single_frame_prefix_encodingtrue►
◄by wuwukasi（bilibili）true►
CategoryWanAnimatePlus
Inputs (25)
| Name | Type | Default | Description |
|---|---|---|---|
| vae | WANVAE | — | |
| width | INT | 83264–8096 | Width of the video to generate. SCAIL-2 inputs are aligned to multiples of 32. |
| height | INT | 48064–8096 | Height of the video to generate. SCAIL-2 inputs are aligned to multiples of 32. |
| num_frames | INT | 811–10000 | Number of frames to generate |
| frame_window_size | INT | 811–10000 | SCAIL-2 chunk window length. Automatically rounds down to 4n+1. Values different from the normalized num_frames enable built-in loop generation with 5-frame handoff; oversized values are clamped to the largest valid window that fits. |
| force_offload | BOOLEAN | true | Offload VAE after encoding to save VRAM |
| pose_strength | FLOAT | 1.0000–10 | Strength of the SCAIL pose stream |
| ref_strength | FLOAT | 1.0000–10 | Strength of the SCAIL reference stream |
| replacement_mode | BOOLEAN | false | False = animation mode (pose mask black bg, reference mask white bg). True = replacement mode (pose mask white bg, reference mask black bg). |
| clip_embedsopt | WANVIDIMAGE_CLIPEMBEDS | Clip vision encoded image | |
| ref_imageopt | IMAGE | Reference image for SCAIL conditioning. If a sequence is connected, only the first frame is used. | |
| bg_imageopt | IMAGE | Optional single background image for animation mode. In single-frame prefix mode it is encoded as an extra background reference latent; in legacy prefix mode it is placed after prefix_frames. Ignored in replacement mode. | |
| pose_imagesopt | IMAGE | Driving pose video. Encoded at half resolution for SCAIL. | |
| prefix_framesopt | IMAGE | Optional prefix images. In single-frame prefix mode these are encoded as reference latents; in legacy mode they hard-freeze the beginning of the canvas. | |
| prefix_maskopt | IMAGE | Optional colored mask images matching prefix_frames. In single-frame prefix mode this follows the reference-mask path; in legacy canvas-prefix mode it is expanded as 1+4+4... and written into the prefix mask frames. | |
| transition_videoopt | IMAGE | Optional transition frames to hard-freeze at the beginning of the canvas. In legacy canvas-prefix mode, transition frames are placed after the prefix frames. | |
| pose_image_maskopt | IMAGE | SCAIL-2 colored per-identity driving pose mask. Background is normalized to black in animation mode and white in replacement mode. | |
| reference_image_maskopt | IMAGE | SCAIL-2 colored per-identity reference mask image. Background is normalized to white in animation mode and black in replacement mode. | |
| tiled_vaeopt | BOOLEAN | false | Use tiled VAE encoding for reduced memory use |
| transition_colormatchopt | COMBO | disabled | Color match transition_video to ref_image. |
| loop_colormatch_referenceopt | COMBO | previous_matched_frame | SCAIL-2 loop color match reference. The first chunk is not color matched when transition_video is not connected. |
| prefix_alpha_cropopt | BOOLEAN | false | Off keeps prefix masks as white-background reference masks in animation mode. On uses black-background masks and alpha-crops prefix_frames. Replacement mode always uses black-background reference masks. |
| preserve_main_ref_backgroundopt | BOOLEAN | true | Animation mode only. Keep the main reference image background. When off, reference_image_mask is normalized to black background and used to alpha-crop ref_image. Ignored in replacement mode. |
| single_frame_prefix_encodingopt | BOOLEAN | true | Encode prefix images as individual full-resolution reference latents instead of expanding the canvas. |
| by wuwukasi（bilibili）opt | BOOLEAN | true | Follow wuwukasi on bilibili |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| image_embeds | WANVIDIMAGE_EMBEDS | — |