Open-source RapidRAW adds 2.5D relighting via depth and normal maps, zero identity loss
zBlackVision11 · reddit · 2026-10-19
Developer Timon implemented 2.5D relighting using depth maps and surface normals in his open-source photo editor RapidRAW (Rust/Tauri/WGPU), as an alternative to diffusion-based relighting.
The problem: Stable Diffusion relighting goes through VAE encoding and img2img, causing identity loss, changed details, or compression artifacts on high-res images.
The approach: no generative model — depth and normals are estimated locally, then lighting is computed directly on original pixels:
- Zero VAE compression; face identity and details stay 100% intact
- Movable point lights, spotlights, and ray-marched shadows in real time
- Full-resolution output without VRAM blowups
Trade-off: won't hallucinate new bounced reflections or change complex backgrounds like diffusion. Free, local, fully open source.
More from coding & agent
- Dev suggests Claude Code compact context before notifying, not after — tonygwu · 2026-10-19
- Air Force recruit open-sources RaiGolmi, a VM workspace for running Claude Code agents unattended for months — BoshBoyBinton · 2026-10-19
- Consultant open-sources agent-first slide editor where Claude Code edits decks from comments — beh_zod · 2026-10-19
- Chat is the wrong shape for AI brainstorming: a dev experiments with tree-shaped idea maps — S_B_B_B_K · 2026-10-19
- Addy Osmani: writing code isn't the bottleneck anymore — build agent self-verification loops — bibryam · 2026-10-19
- Local models are now good enough for first-round code review, dev finds while paying frontier API prices — draginol · 2026-10-19