ComfyUI Node
TS H3 Regenerate Model
Loads the official merged Regenerate LoRA and Sharp branches onto a clean pruned H3 model. Connect the base loader directly, before all generation LoRAs.
TS H3 Regenerate Model
- model
- model
- report
◄lora_name▾►
◄sharp_name▾►
◄reference_checkpoint▾►
◄lora_precisionbf16►
CategoryTS/Video
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | Clean pruned H3 FL2VA model, directly from its loader. No Pika, Turbo or DMAD before this input. | |
| lora_name | COMBO | 1 options: (missing open_h3_regenerate_2k_lora) | |
| sharp_name | COMBO | 1 options: (missing sharp_curve_branches) | |
| reference_checkpoint | COMBO | Only the 32 KB AdaLN reference table is read. This checkpoint is not loaded onto the GPU. | |
| lora_precision | COMBO | bf16 | CPU LoRA factors: about 6.2 GB in bf16 or 12.45 GB in fp32. Release weights are fp32. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| model | MODEL | — |
| report | STRING | — |