ComfyUI Node
Hy-Nomas Start From Image
Starts a world from a picture: the image, held still, becomes segment 0 and the world continues from it with segment 1. Hy-Nomas has no image input; this puts the image in its memory of the past, which later segments continue (not trained for, but it holds the look well). Segment 0 of the script should describe the image; its actions and times are not used. Feed the world to Segment 1 or to Generate World.
Hy-Nomas Start From Image
- model
- clip
- encoder
- script
- image
- world
- summary
◄width832►
◄height480►
◄still_frames3►
◄seed0►
◄history_frames18►
◄int8_kvtrue►
CategoryHy-Nomas
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| clip | CLIP | — | |
| encoder | HYNOMAS_ENCODER | — | |
| script | HYNOMAS_SCRIPT | — | |
| image | IMAGE | Scaled to cover width x height and center-cropped | |
| width | INT | 83232–4096 | Frame size, a multiple of 32. Tencent's defaults: 1280x704 (their 720p; the latent grid is 40x22) or 832x480. Every segment of a world must use the same size. |
| height | INT | 48032–4096 | Frame size, a multiple of 32. Tencent's defaults: 1280x704 (their 720p; the latent grid is 40x22) or 832x480. Every segment of a world must use the same size. |
| still_frames | INT | 31–10 | Latent frames the image is held for (4 video frames each): the model's memory of it. More starts the motion more gently. Decode keeps one of these frames by default (its start_image option). |
| seed | INT | 00–18446744073709550000 | Seed of the cache pass's re-noising |
| history_framesopt | INT | 180–1024 | — |
| int8_kvopt | BOOLEAN | true | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| world | HYNOMAS_WORLD | — |
| summary | STRING | — |