ArtTic-LAB is a hardware-agnostic, node-based generative AI platform built from the ground up for performance and aesthetics.
While originally forged for Intelยฎ Arcโข GPUs, this iteration leverages Native PyTorch implementations to support NVIDIA (CUDA) and AMD (ROCm) with equal proficiency. It moves away from legacy wrappers (like IPEX) in favor of modern, upstream native support and the SDNQ quantization engine for superior memory efficiency.
Whether you are a developer scripting via the CLI or an artist using the infinite node canvas, ArtTic-LAB removes the friction between idea and image.
ArtTic-LAB features a responsive, glassmorphic Node-Based UI. Unlike rigid tabbed interfaces, the infinite canvas allows you to arrange your workspace exactly how you want it.
Weโve packed ArtTic-LAB with features designed to maximize performance on consumer hardware.
| Feature Group | Description |
|---|---|
| Universal Core โ๏ธ | Native XPU/CUDA/ROCm: Auto-detects hardware and utilizes native torch.xpu, torch.cuda, or torch.hip. No heavy external compilation required.PyTorch 2.9+ Ready: Built on the latest nightly builds for maximum stability and speed. |
| Memory Magic ๐ง | SDNQ Integration: Built-in support for sdnq quantization. Run massive models like FLUX.1 on cards with limited VRAM (8GB/12GB) without crashing.Smart VRAM Management: Proactive resolution limiting based on available VRAM and reactive OOM handling that auto-clears cache. |
| Pipeline Mastery ๐ง | Auto-Architecture Detection: Simply load a .safetensors file. The core logic analyzes layer keys to automatically distinguish between SD1.5, SD2.1, SDXL, SD3, and FLUX.Hybrid Loading: Automatically fetches required encoder components from HuggingFace while using your local checkpoint weights. |
| Modern UI/UX โจ | Async Backend: Built on FastAPI and Uvicorn with a WebSocket layer. The UI never freezes during generation. Node Canvas: Draggable, scalable interface nodes for Prompts, Models, Parameters, and Previews. Integrated Gallery: Seamlessly browse, zoom, and manage your creations with embedded metadata reading. |
All demo images generated locally using ArtTic-LAB.
- Miniconda or Miniforge (Python 3.10+ recommended).
- Git installed and available in PATH.
Clone the repository and run the installer script. It will create a Conda environment named ArtTic-LAB and auto-detect your GPU vendor to install the correct PyTorch version.
Windows:
install.batLinux / macOS:
chmod +x install.sh
./install.shStart the server and open the Web UI.
Windows:
start.batLinux / macOS:
./start.shNote: Access the UI at
http://127.0.0.1:7860. To expose a public link via Ngrok, launch withpython app.py --share.
โโโ ๐ArtTic-LAB
โโโ ๐assets
โโโ ๐demos
โโโ 1.png ... 10.png
โโโ ArtTic-LAB-CLI.png
โโโ ArtTic-LAB-GUI-Dark.png
โโโ ArtTic-LAB-GUI-Light.png
โโโ Banner.png
โโโ logo.png
โโโ ๐core
โโโ __init__.py
โโโ logic.py # Main business logic & state management
โโโ metadata_handler.py # PNG Info reader/writer
โโโ prompt_book.py # TOML prompt management
โโโ ๐helpers
โโโ __init__.py
โโโ cli_manager.py # Logging & System Info
โโโ ๐loras
โโโ LORAS.md # Place .safetensors LoRAs here
โโโ ๐models
โโโ CyberRealisticXL.safetensors
โโโ MODELS.md # Place Checkpoints here
โโโ ๐outputs # Generated images saved here
โโโ ๐pipelines # Abstracted Diffusion Pipelines
โโโ __init__.py # Pipeline selector logic
โโโ base_pipeline.py
โโโ flux_pipeline.py
โโโ sd15_pipeline.py
โโโ sd2_pipeline.py
โโโ sd3_pipeline.py
โโโ sdxl_pipeline.py
โโโ ๐tests
โโโ init.py
โโโ run_tests.py
โโโ test_core.py
โโโ test_env.py
โโโ ๐web
โโโ ๐assets
โโโ logo.png
โโโ ๐fonts
โโโ Poppins.ttf
โโโ Silkscreen.ttf
โโโ ๐icons
โโโ material-symbols-outlined.woff2
โโโ material-symbols-rounded.woff2
โโโ material-symbols-sharp.woff2
โโโ __init__.py
โโโ index.html # Main Entry Point
โโโ script.js # Frontend Logic (WebSocket)
โโโ server.py # FastAPI Server
โโโ style.css # Glassmorphic Styles
โโโ .gitignore
โโโ .prettierrc
โโโ app.py # Launcher Script
โโโ ARC-GPU.sh # Intel Arc Diagnostics
โโโ How_It_Works.md # Technical Documentation
โโโ install.bat
โโโ install.ps1
โโโ install.sh
โโโ LICENSE
โโโ prompts.toml # Prompt Database
โโโ README.md
โโโ requirements.txt
โโโ start.bat
โโโ start.ps1
โโโ start.sh
โโโ test.sh
Distributed under the MIT License. See LICENSE for more information.












