ComfyUI Node
Load LDF-VFI Model
A ComfyUI node in video/LDF-VFI with 5 inputs and 1 output.
Load LDF-VFI Model
- model
◄modelonecat-ai/LDF-VFI►
◄tile_size256►
◄tile_overlap64►
◄vae_batch_size8►
◄attention_typeslide_chunk_all_block_2x1x1►
Categoryvideo/LDF-VFI
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | onecat-ai/LDF-VFI | Official LDF-VFI transformer + conditional VAE. Downloads ~6.4 GB on first use and needs about 20 GB VRAM. |
| tile_size | INT | 256128–1024 | Spatial VAE tile size. Larger tiles can improve throughput but use more VRAM. |
| tile_overlap | INT | 648–512 | Spatial overlap blended between VAE tiles. Must be smaller than tile_size. |
| vae_batch_size | INT | 81–32 | VAE temporal-tile batch size. Lower this first if VAE encoding or decoding runs out of VRAM. |
| attention_type | COMBO | slide_chunk_all_block_2x1x1 | Official quick-start sparse attention is recommended. Full attention is extremely memory-intensive. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| model | LDF_VFI_MODEL | — |