Browser ML
Local browser inference with ONNX and WebAssembly
StyleForge
A browser ML project that runs neural style transfer on the user's device with ONNX Runtime and WebAssembly.
ONNX RuntimeWebAssemblyTypeScriptFrontendImage processing
View source How the system fits together.
-
Input
Image selection in the browser
-
Run
ONNX inference through WebAssembly
-
Review
Progress and result comparison
StyleForge runs neural style transfer on the user’s device. It loads an ONNX model through WebAssembly, handles image input and output, reports progress, and keeps the page responsive while inference runs.
What I built
- ONNX Runtime for portable model inference.
- WebAssembly execution in the browser.
- Image interaction with progress feedback.
- Frontend state for model work that takes time.
Engineering notes
Browser inference is slow enough that silence looks broken. StyleForge reports progress and keeps the controls usable until the result is ready to compare.