ComfyUI Node
H3 Inputs
Several MiniMax H3 references on one node: keyframes, reference pictures and reference audio, with slots that grow as you fill them. Each slot names the tag it answers to — picture_2 is <Picture 2>. Reference video goes on H3 Video, which keeps a clip and its soundtrack together. Chain onto other H3 nodes in either direction.
H3 Inputs
- first_frame
- last_frame
- pictures
- audios
- h3_inputs
- h3_inputs
- report
◄sizeauto►
CategoryNynxz/H3
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| size | COMBO | auto | How much detail the reference pictures are fed in at. Reference rows are processed on every sampling step, so 'max' costs several times more. Applies to the pictures here, not to the keyframes, which are canvas-sized. |
| first_frameopt | IMAGE | Opens the clip. Wiring either keyframe selects H3's fl2va task, which cannot be combined with references — H3 Encode's report says what it set aside. | |
| last_frameopt | IMAGE | Closes the clip. Can be used with or without a first frame. | |
| picturesopt | COMFY_AUTOGROW_V3 | — | |
| audiosopt | COMFY_AUTOGROW_V3 | — | |
| h3_inputsopt | NYNXZH3_INPUT | Another H3 node upstream. Its assets come first, then this one's. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| h3_inputs | NYNXZH3_INPUT | — |
| report | STRING | The tag each slot ended up with, counting anything wired upstream. |