Ein sanfterer, verspielter Rhythmus.

Can the website feel as welcoming in motion as it does in a still picture? The full inherited theme palette now eases over 480 ms, including surfaces, text, borders and decorative gradients. The saved/system theme still applies before the first paint; older browsers without registered color properties retain immediate usable changes. Image and 3D previews share a soft 360 ms entrance, 190 ms exit and fading blurred veil. Interruption reverses from the current rendered dialog and backdrop. A stale close cannot hide a reopened preview, and connected triggers regain keyboard focus. Disclosure closure measures its actual collapsed outer box and cleans transient styles after rejection/reversal. Cards, buttons, navigation and source arrows have small lift/press responses. Reduced motion settles active effects and makes changes immediate; decorative Pause retains its ambient-only scope and models still play deliberately. Measured on 16 October 2026: the scoped publication/localization/pages/evidence/export/overview/sync/assets/fonts Python suite passed 165 checks in 35.91 seconds. The actual Node renderer, menu, chronology and new lifecycle handlers passed 28 + 5 + 7 + 12 = 52 checks. Chromium 151 passed 72 checks on frozen preview ux-polish-f76c4e5de31b3e2d, including intermediate body/header/hero colors, sixteen registered palette transitions, rapid reversals, interrupted/reopened dialogs, restored focus, reduced motion, a paused German phone model, six no-JavaScript routes and 48 route/width/theme conditions without horizontal overflow. These are website functional checks, not simulation, performance or full accessibility acceptance. Before frames retain source ux-polish-9143d039a3c83874; first-pass frames retain ux-polish-e2aaff78070ca31f; final frames retain ux-polish-f76c4e5de31b3e2d. Every image and sidecar in session ux-polish-20261016 was individually inspected and approved. The theme evidence pauses actual transitions at 240 of 480 ms; modal evidence pauses the actual entrance at 120 of 360 ms. This exposes an intermediate rendered state rather than fabricating a transition from stills. German phone preview 20261016T141936995000-5d8888b1c27d shows the deliberately loaded, paused Moss model. The first Python run and phone screenshot attempt were interrupted by a full filesystem. The successful reruns followed consolidation of exact duplicate files from this task's own temporary previews and removal of its failed temporary test fixtures; original source snapshots and evidence remain preserved. The test contributor received and read the shared documentation standard. The active house contributor received the standard and explicit website ownership boundaries through the project-authorized coordination channel; house/model/site-selection files were not edited by this increment. Next is to observe the rhythm on physical devices; actual publication and live-byte results remain in the website-sync receipt.