Forge Engine: a native Rust core under Photo Forge
Photo Forge now has a native engine written in Rust for mask and raster operations, connected to Python through PyO3 and backed by a reference fallback.
Illustrative photograph. It does not show Forge software.
Photo Forge's architecture splits responsibilities on purpose. Python with Qt (PySide6) owns the interface, decisions, and the document. A native Forge Engine, written in Rust, does the heavy computation.
What landed
- October 1, 2026: the engine foundation. A Rust crate built with PyO3 and maturin, with a typed bridge into the application.
- October 3, 2026: native mask and raster operations. Mask fill, mask writes, crop, alpha bounds, and applying a mask.
The reference fallback
Every native operation has a Python reference implementation. Tests check that the two agree, and Photo Forge keeps working if the native module is unavailable. That makes the engine safe to grow one operation at a time.
Measured, not assumed
At the time of the October audit the engine had 27 Rust tests and 56 Python tests. The same audit also found two full-canvas operations where the native path was not yet faster than the reference path. That is exactly why a benchmark harness is being built before more optimization work: performance claims should come from measurements.
Next
A native separable blur kernel is in progress. It will speed up Drop Shadow and local filters once it is integrated and measured.
