ComfyDeploy: How ComfyUI-DiffusersImageOutpaint works in ComfyUI?

ComfyUI nodes for outpainting images with diffusers, based on diffusers-image-outpaint by fffiloni.

Updates:

• 17/11/2024:
  • Added more options to Pad Image node (resize image, custom resize image percentage, mask overlap percentage, overlap left/right/top/bottom).
  • Side notes:
    • Now images with round angles work, since the new editable mask covers them, like in the original huggingface space.
    • You can use "mask" and "diffusers outpaint cnet image" outputs to preview mask and image.
    • You can find in the same workflow file the workflow with the checkpoint-loader-simple node and another one with clip + vae loader nodes.
• 22/10/2024:
  • Unet and Controlnet Models Loader using ComfYUI nodes canceled, since I can't find a way to load them properly; more info at the end.
  • Guide to change model used.
• 20/10/2024: No more need to download tokenizers nor text encoders! Now comfyui clip loader works, and you can use your clip models. You can also use the Checkpoint Loader Simple node, to skip the clip selection part.
• 10/2024: You don't need any more the diffusers vae, and can use the extension in low vram mode using sequential_cpu_offload (also thanks to zmwv823) that pushes the vram usage from 8,3 gb down to 6 gb.

To do list to change model used:

• • ✅ ComfyUI Clip Loader Node
• ~⬜ ComfyUI Load Diffusion Model Node~ (more info below)
• ~⬜ ComfyUI Load Conotrolnet Model Node~ (more info below)

Installation

• Download this extension or git clone it in comfyui/custom_nodes, then (if comfyui-manager didn't already install the requirements or you have missing modules), from comfyui virtual env write cd your/path/to/this/extension and pip install -r requirements.txt.
• Download models in comfyui/models/diffusion_models:
  • model_name:
    • unet:
      • diffusion_pytorch_model.fp16.safetensors (example)
      • config.json (example)
    • scheduler:
      • scheduler_config.json (example)
    • model_index.json (example)
  • controlnet_name:
    • config_promax.json (example), diffusion_pytorch_model_promax.safetensors (example)
• (Dual) Clip Loader node: if you use the Clip Loader instead of Checkpoint Loader Simple, and want to use RealVisXL_V5.0_Lightning, it works with clip_I and model.fp16 (from sdxl-base), and sdxl type; you can use this workflow.

Overview

• Minimum VRAM: 6 gb with 1280x720 image, rtx 3060, RealVisXL_V5.0_Lightning, sdxl-vae-fp16-fix, controlnet-union-sdxl-promax using sequential_cpu_offload, otherwise 8,3 gb;
• ~As seen in this issue, images with square corners are required~.

The extension gives 4 nodes:

• Load Diffusion Outpaint Models: a simple node to load diffusion models. You can download them from Huggingface (the extension doesn't download them automatically);
• Paid Image for Diffusers Outpaint: this node resizes the image based on the specified width and height, then resizes it again based on the resize_image percentage, and if possible it will put the mask based on the alignment specified, otherwise it will revert back to the default "middle" alignment;
• Encode Diffusers Outpaint Prompt: self explanatory. Works as clip text encode (prompt), and specifies what to add to the image;
• Diffusers Image Outpaint: This is the main node, that outpaints the image. Currently the generation process is based on fffiloni's one, so you can't reproduce a specific a specific outpaint, and the seed option you see is only used to update the UI and generate a new image. You can specify the amount of steps to generate the image.

You can also pass image and mask to vae encode (for inpainting) node, then pass the latent to a sampler, but controlnets and ip-adapters won't always give good results like with diffusers outpaint, and they require a different workflow, not covered by this extension.

Change model used

• Main model: On huggingface, choose a model from text2image models (sdxl and maybe sd1.5 model types should work, while flux doesn't), then create a new folder named after it in comfyui/models/diffusion_models, then download in it the subfolders unet (if not available use transformer) and scheduler.
  • Hint: sometimes in the unet or transformer folder there are more model files and not all are required. If you have model.fp16 and model, I suggest you to use the fp16 variant; if you have model-001-of-002, model-002-of-002, model, choose model (instead of the fragmented version).
• Controlnet model: download config.json and the safetensors model.

Unet and Controlnet Models Loader using ComfYUI nodes canceled

I can load them but then they don't work in the inference code, since comfyui load diffusers models in a different format (reddit post).

Credits

diffusers-image-outpaint by fffiloni