Sign inSign up

zylarian/dockyard-stable-diffusion-webui

By zylarian

•Updated 10 months ago

Image
0

1.3K

zylarian/dockyard-stable-diffusion-webui repository overview

⁠Stable Diffusion WebUI

AUTOMATIC1111 WebUI for Stable Diffusion - easier interface than ComfyUI.

⁠Quick Start

# Build the image
make service stable-diffusion-webui build

# Start service (CPU mode)
make service stable-diffusion-webui start

# Access
open http://localhost:7860

⁠Features

  • Text-to-image generation
  • Image-to-image
  • Inpainting
  • Outpainting
  • ControlNet support
  • LoRA models
  • Extensions support
  • Multiple models
  • CPU mode (default)

⁠Configuration

Runs in CPU mode by default (no GPU required).

⁠Download Models

Place Stable Diffusion models in ./models/Stable-diffusion/

Popular models:

  • Stable Diffusion v1.5
  • Stable Diffusion v2.1
  • SDXL

Download from:

⁠Ports

  • 7860 - Web interface

⁠Volumes

  • ./models - Stable Diffusion models (safetensors/ckpt)
  • ./outputs - Generated images
  • ./extensions - WebUI extensions
  • ./embeddings - Textual inversion embeddings
  • ./logs - Application logs

⁠Usage

⁠Generate Images
  1. Access http://localhost:7860⁠
  2. Enter prompt
  3. Configure settings (steps, CFG scale, size)
  4. Click Generate
  5. Images saved to ./outputs
⁠Install Extensions
  1. Go to Extensions tab
  2. Available or Install from URL
  3. Apply and restart

⁠GPU Support

To enable GPU (if available):

  1. Edit docker-compose.yml:
runtime: nvidia
environment:
  - COMMANDLINE_ARGS=--listen --port 7860
  1. Restart service

⁠Resource Usage

CPU Mode:

  • CPU: 4 cores max, 2 reserved
  • Memory: 8GB max, 4GB reserved
  • Generation time: 1-5 minutes per image

GPU Mode (if enabled):

  • Significantly faster (seconds)
  • Requires NVIDIA GPU + CUDA

⁠Build

# Local build
cd images/ai/stable-diffusion-webui
./scripts/build.sh

# Multi-arch publish
./scripts/publish.sh

⁠Troubleshooting

⁠Slow generation

Normal in CPU mode. Consider:

  • Reduce image size
  • Reduce sampling steps
  • Use smaller models
⁠Out of memory
  • Reduce batch size
  • Lower image resolution
  • Increase Docker memory limit

⁠Security

  • Not exposed publicly by default
  • No authentication - use firewall

Tag summary

Content type

Image

Digest

sha256:c0f199929…

Size

187.1 MB

Last updated

10 months ago

docker pull zylarian/dockyard-stable-diffusion-webui