ComfyUI Node
arkennemasis Hailuo Scene (one clip, start to finish)
Render ONE scene end to end with Hailuo MiniMax H3 — condition, sample, decode video+audio, mux and write the file — then free everything. Inside a per-scene loop this keeps peak memory at one scene instead of holding every scene's latents at once.
arkennemasis Hailuo Scene (one clip, start to finish)
- model
- clip
- vae
- audio_vae
- sampler
- sigmas
- image
- last_frame
- video
- path
◄prompt►
◄width1280►
◄height736►
◄length243►
◄seed1000►
◄reseed_each_runtrue►
◄filename_prefixnemasis/scene►
Categoryarkennemasis/Video
Inputs (15)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| clip | CLIP | — | |
| vae | VAE | The VIDEO vae. | |
| audio_vae | VAE | The AUDIO vae — this is what gives the clip its soundtrack. | |
| sampler | SAMPLER | — | |
| sigmas | SIGMAS | — | |
| image | IMAGE | This scene's still — the FIRST frame of the clip. | |
| prompt | STRING | — | |
| width | INT | 128032–16384 | — |
| height | INT | 73632–16384 | — |
| length | INT | 2435–2048 | Frames for THIS scene, wired from Scene List so each clip runs its own duration. forceInput on purpose: a widget carrying a link occupies a widgets_values slot and any disagreement about slot counts shifts every later value. |
| seed | INT | 10000–18446744073709550000 | Base seed. It is mixed with THIS scene's prompt, so every scene of a run samples differently instead of all of them sharing one noise pattern. |
| reseed_each_run | BOOLEAN | true | Draw fresh noise on every run. With this off, re-running an unchanged scene reproduces the SAME clip down to the audio - including the same garbled speech, which looks like the scene is locked. Turn it off only when you want to reproduce a specific take exactly. |
| filename_prefix | STRING | nemasis/scene | Relative to ComfyUI's output dir. Each call appends the next counter, so clips land in scene order. |
| last_frameopt | IMAGE | Optional CLOSING frame. Given one, H3 animates the transition from `image` to this, instead of inventing motion from a single still. This is what a first-frame-to-last-frame transformation needs. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| path | STRING | — |