Native App with Tauri for better GPU acceleration #10
Description
First of all, I want to sincerely thank you for building and open-sourcing Kromacut.
What you've created is genuinely impressive. The Beer-Lambert simulation approach, auto-paint logic, and the overall architecture make it one of the most technically solid open-source tools in this space. It’s rare to see this level of mathematical and engineering clarity in a community project.
Your work made it possible for people like me to experiment, extend, and push the idea further — so truly, thank you for that.
Over the past few weeks, I’ve been working on a fork where I tried to expand Kromacut toward a more productized and performance-focused direction. I wanted to share the enhancements with you:
🚀 Enhancements Added in My Fork
1️⃣ Filament Calibration System
- Interactive calibration wizard for accurate Transmission Distance (TD) measurement
- Confidence scoring system (0–1) based on sample consistency
- Guided workflow: Print test strips → Measure → Validate → Save
- Real-time feedback on calibration quality
Impact:
Removes TD guesswork and improves accuracy of Beer-Lambert simulations.
2️⃣ Advanced Optimizer Suite
- Simulated Annealing (global probabilistic optimization)
- Genetic Algorithm (population-based evolutionary optimization)
- Exhaustive search for small filament sets
- Auto mode that selects algorithm based on problem size
- Deterministic seeding for reproducible results
- Result caching to avoid redundant computation
Impact:
Finds optimal filament ordering 10–100× faster than brute force.
3️⃣ Region Weighting System
- Per-pixel importance weighting during optimization
- Prioritize specific regions (center, edge, uniform)
- Useful for portraits (faces) and landscapes (focal areas)
Impact:
Improves color accuracy where it matters most.
4️⃣ Layer-by-Layer Preview Slider
- Interactive height-by-height preview of the print
- Shows filament transitions and blending progression
- Helps detect issues before slicing
Impact:
Better pre-print validation and fewer failed prints.
5️⃣ Native macOS Application
- Full Tauri integration
- Native builds for Apple Silicon & Intel
- Automated CI/CD via GitHub Actions
- DMG installer generation
- Code signing setup
- Improved GPU utilization
Impact:
2–3× faster than the web version with a more polished distribution model.
⚡ Performance & Stability Improvements
6️⃣ Async Mesh Generation
- Periodic yielding during mesh generation
- Prevents UI freezing on large models (>100K pixels)
- Progress tracking for user feedback
7️⃣ Safety Guardrails
- Memory warning when color count exceeds safe threshold
- Prevents empty exports from crashing
I’ve tried to keep the core architectural philosophy intact and build on top of the strong foundation you created.
If you're open to it, I’d love to:
- Discuss potential upstream contributions
- Align on architectural direction
- Or collaborate on future improvements
Again — thank you for building this. It’s a genuinely solid piece of engineering.