ComfyUI Node
Blend Latent Chunks (Pre-Decode)
Join two video chunk latents along temporal dimension before decoding. hard_cut (recommended): clean cut at overlap midpoint — no dissolve. slerp: spherical interpolation preserving latent magnitude. overlap_frames is in LATENT space (Wan: pixel_overlap / 4).
Blend Latent Chunks (Pre-Decode)
- latent_a
- latent_b
- latent
- total_latent_frames
◄overlap_frames4►
◄blend_curvehard_cut►
CategoryVideoChunkTools
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| latent_a | LATENT | — | |
| latent_b | LATENT | — | |
| overlap_frames | INT | 41–64 | Overlap in LATENT temporal frames. For Wan: pixel_overlap / 4 (e.g. 16px = 4 latent) |
| blend_curve | COMBO | hard_cut | hard_cut = clean cut at midpoint (best for rolling reference). slerp = spherical interpolation (preserves latent magnitude). linear/ease_in_out/sigmoid = alpha-blend crossfades. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| latent | LATENT | — |
| total_latent_frames | INT | — |