⚠️ Experimental & UNTESTED ⚠️ AMD and GTX Fix Packs available on discord: https://discord.gg/SYEvvnTt
Welcome to Perchance Revival! This is a user-friendly Gradio web application designed to bring back the experience of the old Perchance image generator by featuring the specific Stable Diffusion 1.5 models and common generation parameters/presets it used.
Generate images effortlessly and for free, directly on your own PC. This version is designed to prioritize and utilize your NVIDIA GPU for significantly faster generation if you have one, while still offering a CPU option for broader compatibility. The goal is to make local SD 1.5 generation as simple as possible, especially for those familiar with the old Perchance site.
This project is primarily designed for Windows users, offering a straightforward setup with easy-to-use batch files. Manual setup options are also provided for other platforms or advanced users.
Please make sure python is installed by downloading the official version, installing it and REMEMBER to tick the small box that says 'add to system path' - just click next/continue and youre done - you can start installing the app: (https://www.python.org/downloads/)
This is the biggest update yet, adding a whole new way to create!
-
Introducing Image-to-Image (img2img)! A whole new tab has been added that lets you upload your own images and transform them with a prompt. Turn your doodles into masterpieces, change the style of a photo, or modify existing AI art.
-
Dedicated Hires. fix Section: Upscale and add significant detail to your creations with a simple, dedicated dropdown menu in the Text-to-Image tab.
-
Increased Image Count: You can now generate up to 12 images in a single batch.
-
Lossless PNG Output: Images now save in PNG format by default.
-
UI/UX Improvements: The UI is now organized with clear tabs, and the image gallery thumbnails are smaller to save screen space.
-
Secondary Style dropdown selection with presets
-
✨ 01/08/2025: More Perchance models and prompts tracked down. Development momentum is back!
-
✨ 11/07/2025: New realistic model - CyberRealistic - added.
-
✨ 11/06/2025: Enhanced Portability & Local Model Caching!
-
Email model suggestions to: [email protected].
⚠️ **beta release notice **
Image-to-Image is a powerful way to use one of your own pictures as a starting point for a new creation.
-
Click the 'Image-to-Image' Tab: Select the tab at the top of the interface.
-
Upload Your Image: Drag and drop an image into the "Input Image" box.
-
Write a Prompt: Describe what you want the final image to look like.
-
Set the 'Strength' Slider: This is the most important setting! It tells the AI how much to change your original image. A low strength (0.1 - 0.4) makes small changes, while a high strength (0.8+) will only use your image as loose inspiration. A medium value is a great place to start.
-
Generate! Click 'Generate from Image' and see what you get!
Hires. fix is a feature in the Text-to-Image tab that creates larger, more detailed images by generating a small image first and then intelligently upscaling it.
- To use it: First, find a seed you like with your base settings. Then, open the 'Hires. fix' menu, tick the checkbox to enable it, and adjust the 'Upscale by' and 'Denoising strength' sliders. Click generate again for a high-resolution result.
CPU Users Please Note: Hires. fix is very computationally intensive. It will be extremely slow on CPU-only systems. It is strongly recommended to only use this feature if you have a compatible GPU.
-
Text-to-Image Mode: Generate art from a simple text description.
-
Image-to-Image Mode: Upload your own image and transform it with a prompt.
-
Device Agnostic: Run on your CPU or leverage your NVIDIA GPU for faster generation.
-
Comprehensive Control: Full control over prompts, steps, CFG scale, schedulers, and seeds.
-
User-Friendly Interface: A clean UI with organized tabs and controls.
-
Safety First (Note): The built-in safety checker is disabled to allow for maximum creative freedom. Please use responsibly.
Generating images on a CPU will always be slower than on a GPU, but you can significantly speed up the process by changing a few settings. If you find the wait times are too long, try these tips:
-
Avoid Hires. fix: As mentioned above, this feature performs a second, intensive pass and will drastically increase generation time on a CPU. It's best left disabled.
-
Lower the Inference Steps: The default is 30, but you can get surprisingly good results with fewer steps. Try starting with 15 steps. With some models and schedulers, you can even get decent images at 10 steps.
-
Use a Fast Scheduler: The Euler scheduler is generally one of the fastest available. Sticking with it is a good way to keep times down.
-
Start with Smaller Images: Generating a 512x512 image will be much faster than generating a 768x1024 one.
-
Experiment! The key is to find a balance between speed and quality that you are happy with. Try generating the same prompt with 10, 15, and 20 steps to see how much of a difference it makes for both the image quality and your wait time.
-
Windows Operating System: The provided batch files (.bat) are for Windows.
-
Python: 3.8 or higher. You can download it from python.org.
-
Hardware:
-
A modern CPU.
-
For GPU speed, a compatible NVIDIA GPU with up-to-date CUDA drivers. At least 6–8GB VRAM is recommended.
-
-
Internet Connection: Required for the initial download of models.
-
Download: Go to the GitHub page, click the green < > Code button, and select "Download ZIP".
-
Extract: Extract the file to a location on your computer.
-
Setup: Run setup-CPU.bat for CPU or setup-GPU.bat for GPU.
-
Wait: A command prompt will open and install everything. Please be patient.
To get the latest features, simply double-click update.bat in the project folder.
Once setup is complete, just double-click the run.bat file. A browser window should automatically open with the app.
-
Clone: Run _git clone https://github.com/Raxephion/Old-Perchance-Revival-WebUI.git_ and then cd Old-Perchance-Revival-WebUI.
-
Create Environment: Run python -m venv venv. Then activate it (venv\Scripts\activate on Windows, source venv/bin/activate on Linux/macOS).
-
Install Dependencies: Run pip install -r requirements.txt.
-
Install PyTorch: Find the correct command for your system at the PyTorch website.
-
Run: Run python main.py.
- Simply delete the project folder. This app is completely portable.
This project is licensed under the MIT License - see the LICENSE file for details.
Feel free to buy me a coffee ;) https://ko-fi.com/raxephion
This project is open-source and free to use, modify, and even break.
But if it does break, please don’t come break me 😉 — use it at your own risk.
Contributions and bug reports are always welcome!