ComfyUI Node
YuE2 LoRAs
Up to four LoRAs for YuE2, each with its strength. A LoRA trained on YuE2's planner shapes the plan - structure, melody, genre -, one trained on its acoustic model the sound; the node reads which from the file. A LoRA that does not fit - another model's, a broken file, weights of another shape - is left out and named in the summary. All slots on none: YuE2 as loaded.
YuE2 LoRAs
- model
- clip
- MODEL
- CLIP
◄lora_1none►
◄strength_11.00►
◄lora_2none►
◄strength_21.00►
◄lora_3none►
◄strength_31.00►
◄lora_4none►
◄strength_41.00►
CategoryPlenio/Model
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | YuE2's MODEL (the acoustic model). | |
| clip | CLIP | YuE2's CLIP (the planner). | |
| lora_1opt | COMBO | none | A YuE2 LoRA from models/loras - for the plan (AR, the CLIP half) or the sound (NAR, the MODEL half): the node reads which from its weights. ComfyUI's layout and m-a-p's original one. |
| strength_1opt | FLOAT | 1.00-4–4 | How strongly it acts: 1 as trained, 0 off. A plan LoRA often sounds best near 0.5, a sound LoRA near 1 - its card says. Far above its trained strength it sounds broken. |
| lora_2opt | COMBO | none | A YuE2 LoRA from models/loras - for the plan (AR, the CLIP half) or the sound (NAR, the MODEL half): the node reads which from its weights. ComfyUI's layout and m-a-p's original one. |
| strength_2opt | FLOAT | 1.00-4–4 | How strongly it acts: 1 as trained, 0 off. A plan LoRA often sounds best near 0.5, a sound LoRA near 1 - its card says. Far above its trained strength it sounds broken. |
| lora_3opt | COMBO | none | A YuE2 LoRA from models/loras - for the plan (AR, the CLIP half) or the sound (NAR, the MODEL half): the node reads which from its weights. ComfyUI's layout and m-a-p's original one. |
| strength_3opt | FLOAT | 1.00-4–4 | How strongly it acts: 1 as trained, 0 off. A plan LoRA often sounds best near 0.5, a sound LoRA near 1 - its card says. Far above its trained strength it sounds broken. |
| lora_4opt | COMBO | none | A YuE2 LoRA from models/loras - for the plan (AR, the CLIP half) or the sound (NAR, the MODEL half): the node reads which from its weights. ComfyUI's layout and m-a-p's original one. |
| strength_4opt | FLOAT | 1.00-4–4 | How strongly it acts: 1 as trained, 0 off. A plan LoRA often sounds best near 0.5, a sound LoRA near 1 - its card says. Far above its trained strength it sounds broken. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| MODEL | MODEL | YuE2's acoustic model with the sound LoRAs. |
| CLIP | CLIP | YuE2's planner with the plan LoRAs. |