ComfyUI Node
Genfocus Model Loader (Components)
A ComfyUI node in Genfocus with 8 inputs and 1 output.
Genfocus Model Loader (Components)
- pipeline
◄transformernone►
◄t5_encodernone►
◄clip_l_encodernone►
◄vaenone►
◄precisionbf16►
◄deblur_loranone►
◄bokeh_loranone►
◄offload_to_cpufalse►
CategoryGenfocus
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| transformer | COMBO | none | FLUX transformer .safetensors. Looks in checkpoints/, unet/, diffusion_models/. Format: 'folder:filename'. Genfocus LoRAs were trained on FLUX.1-dev - fine-tunes will load but quality is untested. Do not use FLUX.1-schnell (timestep distillation differs). |
| t5_encoder | COMBO | none | T5-XXL text encoder .safetensors. Use bf16 or fp16 variants (e.g. t5xxl_fp16.safetensors). fp8 variants are not currently supported by this loader. |
| clip_l_encoder | COMBO | none | CLIP-L text encoder .safetensors (e.g. clip_l.safetensors) |
| vae | COMBO | none | FLUX VAE .safetensors (e.g. ae.safetensors) |
| precision | COMBO | bf16 | Pipeline precision (bf16 recommended for FLUX) |
| deblur_loraopt | COMBO | none | DeblurNet LoRA from models/genfocus/ |
| bokeh_loraopt | COMBO | none | BokehNet LoRA from models/genfocus/ |
| offload_to_cpuopt | BOOLEAN | false | Enable CPU offloading to save VRAM (slower) |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| pipeline | GENFOCUS_PIPELINE | — |