Original-RGB finite woodland-edge study

The three-tree companion painting WOG-A-001701 now has a separate finite alpha study, WOG-A-009287. The complete original and its WOG-A-005855 historical comparison are retained. The accepted archive disposition and every existing approval/goal action remain unchanged. The native 1254 by 1254 image is not resized, cropped, rotated, sharpened or repainted. Every one of its 1,572,516 RGB pixels is copied exactly, including RGB behind alpha zero.

The built-in image_gen tool supplies only an aligned alpha guide; its generated RGB is discarded. An authored 48,732-pixel tree/rock/contact core stays opaque. Independent PNG and Sharp decoders prove zero changed RGB pixels. Ordinary HTMLImageElement/Canvas decoding preserves every opaque RGB pixel and every protected-region RGBA pixel. The fringe has 18,521 partial-alpha pixels; 12,875 have RGB readback differences with maximum channel difference 14. These remain explicit. There are no visible image-border pixels. Exact prompt generation.json, native proof native-matte.json, browser proof browser-native-proof.json.

The complete source and complete visible alpha domain inside the native comparison were personally inspected. All three canopies and trunks, the central rock and painted contact/shadow areas remain recognizable. The outer repeated field is removed from the visible domain. The irregular ground/foliage perimeter still needs normal terrain and shadow review; source pixels alone do not prove correct compositing. This is one finite cluster, never a repeated mesh-ground tile.

The linked Woodland Waystone Sanctuary intentions cover forest, hills and grasslands only. No parent canonical scale, anchor, source permission, lifecycle or gameplay binding is inherited. Normal Region/Hex desktop and phone contact, leaf edge, neighbor/UI, seasonal/state, lifecycle, performance and actual production-world appearance remain pending. This static proof adds zero normal world frames, no harvestable forest, no new vegetation state and no player action.

Keep every original, selected native study and the decisive native comparison native-comparison.png. Retire the raw generated alpha input only after final preservation and Atlas checks. Keep the existing ninety-contract world run and preceding lodge release separate.