Accessibility

What works

• The editor is keyboard-operable. Move through the effect palette, add and reorder effects, open and adjust every parameter, and reach the menus, workspace tabs and export controls without a mouse. Sliders respond to arrow keys — hold Shift for larger steps, Home and End for the extremes.
• Controls are named. Every button, slider, dropdown and field reports what it is to a screen reader, including the ones inside panels that start collapsed.
• The page has proper structure — landmarks, a heading order, and a skip link to jump straight to the editor.
• The canvas is labelled, so its presence and purpose are announced.
• Zoom is not blocked. Pinch-zoom and browser zoom both work; you can scale the interface.

What doesn't, and why I'm telling you

• The canvas can’t be described. It shows a live, GPU-rendered image. A screen reader can announce that the canvas is there, but not what your photo now looks like. There is no honest text alternative for “your photo, relit”, and I’m not going to invent a caption that pretends otherwise.
• Some controls are mouse-only. The corner-pin and crop handles, the layer-reorder grips, the 2D colour and light pads, and a few vertical faders. These need real keyboard equivalents designed rather than a focus ring added, and that work isn’t done.
• Five effect-group labels don’t meet contrast guidance.  They’re small text set in their group’s colour, and they measure below the recommended ratio. That colour coding is part of how the app is organised, and I’ve chosen to keep it rather than flatten it — so this is a deliberate exception, not an oversight. Every group is also identified by its name and its position, never by colour alone.

I treat this as work in progress, not a box to tick. If something here blocks you, tell me — and it goes on the list with everything else.