MetricMogul's picture
Update README.md
26d1c37 verified
|
Raw
History Blame Contribute Delete
1.93 kB
---
title: SD15 Light Inpaint CPU
emoji: 🐨
colorFrom: gray
colorTo: gray
sdk: gradio
sdk_version: 6.14.0
python_version: '3.13'
app_file: app.py
pinned: false
license: apache-2.0
short_description: Lightweight SD1.5 inpainting lab for quick CPU-friendly imag
---
# SD15 Light Inpaint CPU
A lightweight SD1.5 inpainting Space for quick masked image edits.
Upload an image, draw over the area you want to repaint, write a prompt, and generate a fast draft.
This Space is a lighter CPU-oriented version of the SD1.5 inpainting workflow. The goal is not maximum quality, but a small and practical image-editing lab that can run with modest resources.
## What it does
- Upload an image
- Draw a mask over the area to change
- Describe what should appear in the masked area
- Generate a quick inpainting result
## Current model stack
- Base model: `stable-diffusion-v1-5/stable-diffusion-inpainting`
- Speed adapter: `latent-consistency/lcm-lora-sdv1-5`
- Scheduler: `LCMScheduler`
- Tiny VAE: `madebyollin/taesd`
- UI: Gradio
## Default settings
The default settings are intentionally conservative for CPU use:
- Max side: `384 px`
- Steps: `4`
- Guidance scale: `1.5`
- Strength: `0.85`
- Batch size: `1`
Higher resolutions and more steps may improve quality, but they will be slower on CPU.
## What it is for
This Space is useful for:
- quick masked edits
- inpainting experiments
- testing low-step SD1.5 generation
- comparing lightweight CPU settings
- building a small image-editing workflow before adding heavier adapters
## What it is not
This is not a production image editor.
It is not a high-resolution SDXL or FLUX replacement.
It is not a face swap tool.
It is not the full IP-Adapter version yet.
The first version focuses on the simplest useful workflow:
```text
image + mask + prompt → inpaint
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference