ComfyUI-Switti

Custom nodes for running the Switti text-to-image model inside ComfyUI.

Nodes

• Load Switti Checkpoint
• Load Switti VAE Checkpoint
• Load Switti Dual CLIP
• Build Switti Pipeline
• Switti Sampler

Required Models

You must provide three sets of weights:

1. Switti core checkpoint (.safetensors)
2. Switti VQVAE checkpoint (.safetensors)
3. Two text encoders:
   • clip-vit-large-patch14
   • CLIP-ViT-bigG-14-laion2B-39B-b160k

The CLIP entries can be folders placed in ComfyUI/models/text_encoders or .safetensors files with matching names in that same folder.

Workflow

Load workflows/switti_basic.json in ComfyUI and fill in:

• Switti core checkpoint (diffusion_models folder)
• Switti VAE checkpoint (vae folder)
• CLIP encoders (text_encoders folder)

Make sure core and VAE resolutions match (512 or 1024).

Notes

• Standard SD VAEs and SDXL CLIP weights are not compatible with Switti.
• If CLIP inputs are swapped, you will get a size mismatch error.