ComfyUI Node
MiniMax H3 Upscaler
MiniMax H3 Upscaler by SOLRICKS: latent enlargement and windowed detail refinement using existing model files.
MiniMax H3 Upscaler
- model
- vae
- images
- conditioning
- audio
- audio_vae
- images
- audio
◄upscale_model▾►
◄megapixels2.00►
◄loraNone►
◄enable temporal chunkingtrue►
◄enable spatial tilingtrue►
◄fps—►
CategoryCustomPostProcess
Inputs (12)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | Plain MiniMax H3 FL2VA diffusion model. | |
| vae | VAE | MiniMax H3 video VAE with a decoder. | |
| images | IMAGE | — | |
| upscale_model | COMBO | 1 options: minimax_h3_latent_upscaler_3d_conv_v1_fp16.safetensors | |
| megapixels | FLOAT | 2.000.1–16 | — |
| lora | COMBO | None | 1 options: None |
| enable temporal chunking | BOOLEAN | true | Allow overlapping time windows for long clips or limited VRAM. |
| enable spatial tiling | BOOLEAN | true | Allow overlapping image tiles when a frame does not fit in VRAM. |
| conditioningopt | CONDITIONING | Optional H3 text prompt describing the clip. | |
| audioopt | AUDIO | Original soundtrack; returned unchanged. | |
| fpsopt | FLOAT | Source frame rate; defaults to 24. | |
| audio_vaeopt | VAE | H3 audio VAE for soundtrack conditioning. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| audio | AUDIO | — |