🎨 LoRA × Sampler × Scheduler Grid Forge — Triple Strike Against Chaos in Generation!
🎨 LoRA × Sampler × Scheduler Grid Forge — Triple Strike Against Chaos in Generation!
Ever found yourself testing a LoRA, switching samplers, tweaking schedulers — and a couple of hours later you can’t remember which exact combo gave you that perfect shot? LoRA × Sampler × Scheduler Grid Forge turns that chaos into neat, labeled grids where everything is clear at a glance.
It’s the tool that saves you hours of testing and makes your experiments reproducible and easy to understand.
🌟 What Makes This Script Special
💡 Triple comparison in a single frame — LoRA weight × Sampler × Scheduler. No more juggling dozens of combinations in your head.
⚡ Two modes of operation:
XY Grid — automatically iterates through all combinations you’ve selected.
Batch Grid — manually assemble custom sets when you need targeted checks.
🖋 On-grid labels — with a custom font so everything looks clean and on-brand.
🛡 Fallback images — instead of empty cells on errors, you get a readable message, not a hole in your grid.
🧹 Duplicate filtering — no wasted generation or accidental repeats.
📏 Flexible parameter control — from prompt and seed to dimensions, padding, and output format (PNG/WEBP).
📦 Individual cell saving — perfect for detailed analysis and presentation prep.
🎯 Who This Is a Must-Have For
Artists and art directors who want to quickly find “golden” combinations without folder chaos.
LoRA creators who need to see how weight and parameters affect style and anatomy.
Teams that need to share results in a clear, reproducible way.
🚀 How It Speeds Up Your Workflow
Time saver — one run instead of dozens of scattered tests.
Clear documentation — the grid itself is ready for an article, post, or report.
Quality control — easily spot where a model “drifts” or stays consistent.
💬 "It’s like having your own visual microscope for generation — only instead of lenses, you’ve got LoRA, sampler, and scheduler."
⚙️ Installation
cd /extensions/
git clone https://github.com/PupaBoo/lora_x_sampler_x_scheduler_grid_forge
cd lora_x_sampler_x_scheduler_grid_forge
pip install -r requirements.txt(Optional) Add the Barlow-SemiBold.ttf font to the fonts/ folder for improved label display.
🩵 I uploaded the script directly to this article for convenience. Use and enjoy it! 🩵
🖥 How to Use
Launch Forge WebUI and select LoRA × Sampler × Scheduler Grid Forge from the
txt2imgscript dropdown.Choose a mode:
XY Grid — set LoRA weights, samplers, and schedulers for the axes.
Batch Grid — add combinations manually or via the interface.
Configure parameters (prompt, seed, steps, CFG, dimensions, padding, format).
Enable labels and individual cell saving if needed.
Click Generate and watch progress in the terminal.
📁 Output Structure
outputs/
└── 🧪 LoRa x Sampler x Scheduler Grid (Forge)/
├── cells/ # Individual cell images
│ ├── LoRAName_W0.50_Sampler_Scheduler_0_0.webp
│ └── ...
├── xy_grid_1700000000_001.webp # XY Grid output (WEBP format)
├── xy_grid_1700000000_001.png # XY Grid output (PNG format)
├── batch_grid_1700000000_001.webp # Batch Grid output (WEBP format)
└── batch_grid_1700000000_001.png # Batch Grid output (PNG format)🛠 Requirements
Stable Diffusion Forge
Python 3.7+
Libraries:
gradio,Pillow(included with Forge)
⚡ Technical Features
Auto-Correction: Intelligent name matching for samplers and schedulers.
Validation: Comprehensive configuration validation with helpful error messages.
Unicode Support: Proper handling of special characters and emojis in names.
Font Caching: Optimized performance through font caching mechanisms.
Memory Management: Efficient handling of large grids with auto-scaling.
⚠️ Limitations
Mode Restriction: Works exclusively in txt2img mode.
Resource Intensive: Large grids may require significant VRAM and processing time.
Font Dependency: Optimal text rendering requires Barlow-SemiBold.ttf availability.
Grid Size: Very large combinations may produce extremely large output files.
📜 License
MIT License © 2025 - Free to use with attribution.
🙌 Acknowledgments
Built with Gradio and Pillow libraries
Developed with AI assistance tools
Inspired by the needs of the Stable Diffusion artist community
🤝 Contributing
Feedback, bug reports, and feature suggestions are welcome. Please report issues with detailed descriptions of the problem and your configuration.