FastSD CPU

Processor : Intel(R) Xeon(R) Platinum 8375C CPU @ 2.90GHz

Mode

Current working mode

LCM LCM-LoRA LCM-OpenVINO GGUF

Textbox

Negative prompt (Works in LCM-LoRA mode, set guidance > 1.0) :

Generated images

Init image

Textbox

Negative prompt (Works in LCM-LoRA mode, set guidance > 1.0):

Strength

0.1 1

Generated images

Init image

Variations Strength

0.1 1

Generated images

Image

Upscale Mode (2x) | AURA-SR v2 (4x)

Select upscale method(For SD Scale GGUF mode is not supported)

EDSR SD AURA-SR

Generated images

Inference Steps

1 25

Image Height

256 1024

Image Width

256 1024

Number of images to generate

1 50

Guidance Scale

1 10

CLIP Skip

1 12

Token Merging

0 1

Seed

0 999999999

Use seed

Use Safety Checker

Use Tiny AutoEncoder (Fast, moderate quality)

Use locally cached model or downloaded model folder(offline)

Output image format

PNG JPEG

LCM model

Diffusers LCM model ID

LCM LoRA model

Diffusers LCM LoRA model ID

LCM LoRA base model

Diffusers LCM LoRA base model ID

LCM OpenVINO model

OpenVINO LCM-LoRA fused model ID

GGUF diffusion model

GGUF CLIP model

GGUF T5-XXL model

GGUF VAE model

Download and place your LoRA model weights in lora_models folders and restart App

LoRA model

LoRA model weight to load (You can use Lora models from Civitai or Hugging Face .safetensors format)

Initial Lora weight

0 1

Loaded LoRA models

Lora Name

LoRA weight

0 1

Lora Name

LoRA weight

0 1

Lora Name

LoRA weight

0 1

Lora Name

LoRA weight

0 1

Lora Name

LoRA weight

0 1

Download ControlNet v1.1 model from ControlNet v1.1 (723 MB files) and place it in controlnet_models folder,restart the app

Enable ControlNet

ControlNet model

ControlNet model to load (.safetensors format)

ControlNet conditioning scale

0 1

Control image

Preprocessor

Select the preprocessor for the control image

Canny Depth LineArt MLSD NormalBAE Pose SoftEdge Shuffle None

Built with Gradio logo