sd-webui-deoldify

Introduction to DeOldify for Stable Diffusion WebUI

DeOldify for Stable Diffusion WebUI is an innovative extension designed to breathe new life into old photographs and videos by adding vibrant colors to them. This tool is an extension of the AUTOMATIC1111 Stable Diffusion web-ui, which is known for harnessing the powerful capabilities of the DeOldify library, a machine learning model specifically crafted for colorization tasks. This project is offered by SpenserCai and it blends advanced algorithms with user-friendly accessibility, enabling users to easily rejuvenate historical and personal visual media.

Key Features and Updates

DeOldify for Stable Diffusion WebUI continuously evolves, with significant updates enhancing its functionality:

• Open Source DeoldifyBot: As of August 12, 2023, users can access the source code to deploy the DeoldifyBot independently.
• Discord Integration: Released on August 11, 2023, users can now colorize images directly via Discord.
• Video Colorization Support: Announced on August 7, 2023, this feature allows users to apply color to old videos, alongside photos.
• Simple Installation: On August 5, 2023, an easy installation option through the Extensions list was introduced.
• Initial Launch: The project made its debut on August 4, 2023.

Compatibility

DeOldify for Stable Diffusion WebUI is compatible with various systems and Pytorch versions:

Operating Systems

Pytorch Versions

Users are encouraged to contribute by testing other systems or Pytorch versions, and share successful results.

Installation Guide

To install the DeOldify extension:

1. Navigate to the "Extensions" tab in the Stable Diffusion web-ui.
2. Enter the URL https://github.com/SpenserCai/sd-webui-deoldify in the "install from URL" section.

For users with network issues, the extension can be downloaded from Hugging Face Model.

Usage Instructions

Photo Colorization

1. Access the "Extra" tab in the web-ui.
2. Select the "DeOldify" checkbox.
3. Upload the old photograph you wish to colorize.

Application Scenario

By combining DeOldify with tools like Upscale and GFPGAN, users can achieve impressive restoration effects. Here’s a before and after comparison:

Before	After

Video Colorization

To colorize videos:

• Ensure ffmpeg is installed using sudo apt install ffmpeg.
• In the "DeOldify" tab, upload the video, set the "Render Factor," and click "Run".

DeOldifyBot

DeOldifyBot extends the capabilities of DeOldify to real-time communication platforms:

Using DeOldifyBot on Discord

1. Add the bot to your Discord server.

Building Your Own DeOldifyBot

For those interested in building the bot locally:

# Install Golang
# Follow instructions at https://golang.org/doc/install

# Navigate to bot directory
cd bot
# Create release directory
mkdir release
# Build the project
bash build.sh
# Set up configuration
cp ./config.example ./release/config.json
# Edit config.json
# Input your Discord bot token and sd-webui address

# Run the bot
release/DeOldifyBot

Future Development

The team is working on enhancing DeOldify's capabilities, including:

• Improving the speed of video processing.
• Adding repair options for greater image restoration.
• Supporting simultaneous image generation with different settings through scripts.
• Removing the need to add --disable-safe-unpickle at startup.

The project's roadmap ensures continued innovation, making DeOldify a comprehensive tool for creative restoration projects.