Type: huggingface.image_to_image.StableDiffusionInpaint
Namespace: huggingface.image_to_image
Description
Performs inpainting on images using Stable Diffusion. image, inpainting, SD
Use cases:
- Remove unwanted objects from images
- Fill in missing parts of images
- Modify specific areas of images while preserving the rest
Properties
| Property | Type | Description | Default |
|---|---|---|---|
| model | hf.stable_diffusion |
The model to use for image generation. | {'type': 'hf.stable_diffusion', 'repo_id': '', 'path': None, 'variant': None, 'allow_patterns': None, 'ignore_patterns': None} |
| variant | Enum['default', 'fp16', 'fp32', 'bf16'] |
The variant of the model to use for Image-to-Image generation. | fp16 |
| prompt | str |
The prompt for image generation. | `` |
| negative_prompt | str |
The negative prompt to guide what should not appear in the generated image. | `` |
| seed | int |
Seed for the random number generator. Use -1 for a random seed. | -1 |
| num_inference_steps | int |
Number of denoising steps. | 25 |
| guidance_scale | float |
Guidance scale for generation. | 7.5 |
| scheduler | Enum['DPMSolverSDEScheduler', 'EulerDiscreteScheduler', 'LMSDiscreteScheduler', 'DDIMScheduler', 'DDPMScheduler', 'HeunDiscreteScheduler', 'DPMSolverMultistepScheduler', 'DEISMultistepScheduler', 'PNDMScheduler', 'EulerAncestralDiscreteScheduler', 'UniPCMultistepScheduler', 'KDPM2DiscreteScheduler', 'DPMSolverSinglestepScheduler', 'KDPM2AncestralDiscreteScheduler'] |
The scheduler to use for the diffusion process. | EulerDiscreteScheduler |
| loras | List[hf.lora_sd_config] |
The LoRA models to use for image processing | [] |
| ip_adapter_model | hf.ip_adapter |
The IP adapter model to use for image processing | {'type': 'hf.ip_adapter', 'repo_id': '', 'path': None, 'variant': None, 'allow_patterns': None, 'ignore_patterns': None} |
| ip_adapter_image | image |
When provided the image will be fed into the IP adapter | {'type': 'image', 'uri': '', 'asset_id': None, 'data': None} |
| ip_adapter_scale | float |
The strength of the IP adapter | 0.5 |
| pag_scale | float |
Scale of the Perturbed-Attention Guidance applied to the image. | 3.0 |
| latents | torch_tensor |
Optional initial latents to start generation from. | {'type': 'torch_tensor', 'value': None, 'dtype': '<i8', 'shape': [1]} |
| enable_attention_slicing | bool |
Enable attention slicing for the pipeline. This can reduce VRAM usage. | True |
| enable_tiling | bool |
Enable tiling for the VAE. This can reduce VRAM usage. | True |
| enable_cpu_offload | bool |
Enable CPU offload for the pipeline. This can reduce VRAM usage. | True |
| output_type | Enum['Image', 'Latent'] |
The type of output to generate. | Image |
| init_image | image |
The initial image to be inpainted. | {'type': 'image', 'uri': '', 'asset_id': None, 'data': None} |
| mask_image | image |
The mask image indicating areas to be inpainted. | {'type': 'image', 'uri': '', 'asset_id': None, 'data': None} |
| strength | float |
Strength for inpainting. Higher values allow for more deviation from the original image. | 0.8 |
Outputs
| Output | Type | Description |
|---|---|---|
| image | image |
|
| latent | torch_tensor |
Metadata
Related Nodes
Browse other nodes in the huggingface.image_to_image namespace.