ComfyUI Node
H3 Video
Wired video frames as an H3 reference, with an optional soundtrack. Set fps to the frames' real rate — H3 presents reference video at 24 fps with timestamps derived from it. Chains onto H3 Inputs.
H3 Video
- frames
- audio
- latent
- h3_inputs
- h3_inputs
◄fps24.00►
◄detailfull►
CategoryNynxz/H3
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| frames | IMAGE | The video as an IMAGE batch, in order. | |
| fps | FLOAT | 24.001–240 | The frames' actual frame rate. They are resampled onto H3's 24 fps from this, so a wrong value plays the reference at the wrong speed and mislabels its <T.T seconds> timestamps. |
| detail | COMBO | full | How much resolution this reference is fed in at — a pure cost dial. Reference rows ride every sampling step, and a video reference is usually a third of the whole sequence, so 'half' takes roughly a quarter off the run and a similar amount off peak VRAM. It costs no alignment: H3's spatial rope depends on aspect ratio alone, so a half-resolution reference sits in the identical coordinate box and stays registered to the target. Use 'full' when the reference is the only source of fine detail. |
| audioopt | AUDIO | Soundtrack for these frames. Packs the reference as video+audio, and needs H3's audio VAE on the encode node. | |
| latentopt | LATENT | This same video as an H3 latent, when you already have one — the output of an earlier sampler. H3 Encode then skips the VAE encode and uses it directly, which is what a regenerate pass wants: the first pass's own latent, with no decode/re-encode round trip to lose detail through. The frames are still needed, because Qwen3-VL reads pixels. | |
| h3_inputsopt | NYNXZH3_INPUT | Another H3 node upstream. Its assets come first, then this one's. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| h3_inputs | NYNXZH3_INPUT | — |