The ModelSamplingSD3 node in the official workflow sample produces pure noise. After trying ModelSamplingAuraFlow with a value of 5.0, it produced the same σ curve in the Model Preview Override as it did without it, so I removed the flow shift from the workflow (it appears to be shifted by default).
MOSS-Image Prompt Weight is used instead of CLIP NegPip because it produces better results. You can use positive weights to emphasize elements and negative weights to exclude or suppress them.
The sample image was generated using a layout prompt, and these settings are saved in the workflow. Clicking run would recreate the image with the right model.
Will add more features when available.
Basic MOSS t2i workflow
Features:
Quick bypass access
No messy connections. Clean and tidy. Easy to understand and customize.
Prompt Weight
Layout Prompting
Tiled Diffusion for high upscale factors.
Split High/Low noise which is useful when style lora impacts composition
Custom nodes used:
Use Install via Git URL if you can't find it via Node Manager. (in legacy UI, enable with --enable-manager-legacy-ui flag)
rgthree-comfy: https://github.com/rgthree/rgthree-comfy
ComfyUI-KJNodes: https://github.com/kijai/ComfyUI-KJNodes
ComfyUI Impact Pack: https://github.com/ltdrdata/ComfyUI-Impact-Pack
ComfyUI-Easy-Use: https://github.com/yolain/ComfyUI-Easy-Use
ComfyUI-Lora-Manager: https://github.com/willmiao/ComfyUI-Lora-Manager
ComfyUI_essentials: https://github.com/cubiq/ComfyUI_essentials
ComfyUI-mxToolkit: https://github.com/Smirnov75/ComfyUI-mxToolkit
ComfyUI Image Saver: https://github.com/alexopus/ComfyUI-Image-Saver
cg-image-filter: https://github.com/chrisgoringe/cg-image-filter
ComfyUI-FBCNN: https://github.com/Miosp/ComfyUI-FBCNN
ComfyUI-MooshieTiledDiffusion: https://github.com/Mooshieblob1/ComfyUI-MooshieTiledDiffusion
ComfyUI Nvidia VFX Nodes: https://github.com/Comfy-Org/ComfyUI_Nvidia_VFX_Nodes
comfyui-upscale-by-model: https://github.com/cdzcj/comfyui-upscale-by-model (Node Manager links to wrong repository.)