ComfyUI Node
H3 Video Frames (lean)
Get Video Components without the RAM bill: start, duration, a 24 fps re-time and a resize are applied while decoding, so skipped frames are never held, and the frames are filled into the IMAGE once (peak about 1.25x the output).
H3 Video Frames (lean)
- video
- images
- audio
- fps
- frame_count
- video
◄start_seconds0.00►
◄duration_seconds0.0►
◄frame_ratesource►
◄resizenone►
◄resize_value1.00►
◄snap_to_h3true►
CategoryPlagueKind/minimax/upscaler
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| video | VIDEO | From Load Video (or any VIDEO). | |
| start_seconds | FLOAT | 0.000–86400 | Skip this much of the video first. |
| duration_seconds | FLOAT | 0.00–86400 | How much to take from start_seconds on. 0 = to the end. With snap_to_h3 the frame count is rounded to the nearest length H3 uses natively (30 s at 24 fps = 719 frames); the log says what you got. |
| frame_rate | COMBO | source | source = every frame. 24 = re-timed to H3's own rate while decoding (nearest frame: dropped above 24, repeated below); the duration and the soundtrack's sync are kept. |
| resize | COMBO | none | Shrink while decoding, aspect kept, never enlarging: to resize_value megapixels, or the long edge to resize_value px. |
| resize_value | FLOAT | 1.000–16384 | Megapixels (resize = megapixels) or pixels (resize = long edge). |
| snap_to_h3 | BOOLEAN | true | Round duration_seconds to the nearest frame count H3 uses natively (17n + 5), so nothing is padded. Off = exactly seconds x fps. |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| audio | AUDIO | — |
| fps | FLOAT | — |
| frame_count | INT | — |
| video | VIDEO | — |