Lore
Juggernaut XL checkpoint and the image-to-image sampler graph
The main model block of Blender to ComfyUI SDXL image-to-image render pipeline. It uses Juggernaut XL, a Stable Diffusion XL 1.0 checkpoint from CivitAI.
Wiring
- The checkpoint loader outputs a model, a CLIP and a VAE.
- CLIP encodes the positive prompt (from SDXL Prompt Styler (preset styles instead of hand-typed phrases)) and a negative prompt.
- The model passes through the LoRA and ControlNet blocks (multi-controlnet-stack-strength-tuning) to the K sampler.
- The sampler works from the encoded input image, which is what makes it image-to-image rather than text-to-image (img2img-denoise-tuning, ksampler-cfg-steps-seed).
- The VAE decodes the sampler's latent into the final image.
Why this checkpoint
It is an SDXL model, so it can use the existing SDXL LoRAs and ControlNets (base-model-choice-by-addon-ecosystem). Use the checkpoint's recommended CFG and settings as the starting point.