ComfyUI Node
Spellcaster Loader (Auto-Arch)
Auto-detect model architecture and load the full stack. Handles checkpoint (SD1.5/SDXL) and separate-loader (Flux/Klein/Chroma) models transparently.
Spellcaster Loader (Auto-Arch)
- model
- clip
- vae
- arch_key
◄model_name▾►
◄arch_overrideauto►
◄clip_overrideauto►
◄vae_overrideauto►
◄weight_dtypedefault►
CategorySpellcaster
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | COMBO | Model file — architecture auto-detected from filename | |
| arch_overrideopt | COMBO | auto | Override auto-detection |
| clip_overrideopt | COMBO | auto | Override CLIP file (auto = use architecture default) |
| vae_overrideopt | COMBO | auto | Override VAE file (auto = use architecture default) |
| weight_dtypeopt | COMBO | default | Weight data type for UNET models |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| model | MODEL | Diffusion model (UNET) for denoising latents. |
| clip | CLIP | CLIP text encoder for prompt encoding. |
| vae | VAE | VAE for encoding/decoding images ↔ latent space. |
| arch_key | STRING | Detected architecture key (pass to Sampler / Prompt Enhance). |