mlfarinha

mlfarinha / pixlrelight

Public

PIXLRelight: Controllable Relighting via Intrinsic Conditioning

49
5
89% credibility
Found Jun 01, 2026 at 49 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

PIXLRelight is an AI-powered image relighting tool developed by researchers at Oxford University that transforms photographs by applying new lighting conditions, either from reference images or 3D render passes, in under a tenth of a second while preserving fine image details.

How It Works

1
🔍 Discover the relighting tool

You hear about a research tool from Oxford that can change the lighting in any photograph with just one click.

2
📦 Install the tool

You download and set up the program on your computer, and the AI model downloads automatically when you first run it.

3
🖼️ Choose your source photo

You pick any photograph you want to relight - it could be an indoor room, a portrait, or an outdoor scene.

4
Choose your target lighting
📸
Use a reference photo

Show the AI another photograph of a scene with your desired lighting, and let it figure out the lighting from that image automatically.

🎬
Use 3D render passes

If you have 3D modeling experience, you can provide detailed lighting information from a 3D scene you created in Blender.

5
✨ Watch the magic happen

The AI takes your source photo and transforms it, applying the new lighting while keeping all the original details and textures intact.

🎉 Get your relit image

Your transformed image appears in seconds, showing the same scene bathed in completely new light - brighter, darker, warmer, or from a different direction.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 49 to 49 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is pixlrelight?

PIXLRelight is a Python library that relights photographs under new lighting conditions in under a second. You feed it a source image and either a reference photo of your target lighting or a set of Blender Cycles render passes, and it outputs the scene relit under that illumination. The model works by decomposing both images into intrinsic components (albedo, shading, residuals) and using those as conditioning signals for a transformer-based renderer. It preserves fine detail by applying per-pixel gain and bias modulation at the source's native resolution. Weights are automatically downloaded from Hugging Face on first run.

Why is it gaining traction?

The killer feature is the dual-mode conditioning. If you just have a photo of a nice sunset, Mode 1 decomposes it on-the-fly and applies that lighting to your scene. But if you need precise, physically-based control, Mode 2 lets you specify exact PBR lights in Blender and get predictable, controllable results. The feed-forward architecture means no iterative optimization or diffusion sampling -- you get results in a single forward pass. The intrinsic conditioning bridge between real photographs and PBR renders is genuinely novel, solving the classic gap between synthetic and real imagery.

Who should use this?

Game artists and VFX professionals who need to match product photography to concept art lighting without reshoots. Architectural visualization studios wanting to preview designs under different natural lighting conditions. Researchers in intrinsic image decomposition or relighting who need a strong baseline. Anyone building tools that require fast, controllable scene relighting without the overhead of full path tracing.

Verdict

This is a credible academic project from Oxford's PIXL lab with a published paper, but the 49 stars and early-stage maturity mean you should expect rough edges. The documentation is solid for a research release, inference works out of the box, and the dual-mode approach is genuinely useful. The 0.9% credibility score reflects legitimate provenance but limited community validation. Worth evaluating for production use if your pipeline can handle research-grade tooling, but benchmark carefully against your specific use case before committing.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.