You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
charlie — Effect Symbols: define a reusable effect once and link any number of effects to it; editing one propagates to all linked effects. Symbols round-trip through .xsq with a "Convert All Symbols to Effects" escape hatch for compatibility export (#2671)
scott — Windows: experimental Vulkan preview rendering backend (model preview, sequencer grid, waveform, color canvases), switchable via Preferences > Other > "Preview graphics"; defaults to OpenGL with automatic fallback
dan — Linux: Vulkan preview rendering backend with X11 and native Wayland surface support; new "Auto" default picks Vulkan only when OpenGL would fall back to software rendering (llvmpipe)
dkulp — Linux/Windows: GPU-accelerated rendering (Vulkan) for blur, rotozoom, transitions, layer blending and most effects, mirroring the Metal backend on macOS; enable via Preferences > Other > GPU rendering
dkulp — macOS/iPad: Shader effect now renders natively on Metal (GLSL translated to Metal via SPIR-V and cached); set XL_NO_NATIVE_SHADER=1 to force the OpenGL fallback
dkulp/scott — Linux/Windows: Shader effect now renders natively on Vulkan (GLSL -> SPIR-V via glslang); falls back to OpenGL when Vulkan is unavailable
dkulp — GPU/SIMD rendering for additional effects: Tree, Shimmer, Candle (per-node), Wave, Garlands, Fill, Life, Twinkle, and Meteors
dkulp — Galaxy effect: blend passes now scan only the bounding box instead of the whole buffer, up to ~5x faster on large 2D models
dkulp — Much faster saving of large .fseq files via multi-threaded zstd compression
dkulp — Add headless render mode (--headless -s <seq.xsq>): renders sequences to .fseq with no window, then exits
cybercop23 — Effect Presets track unsaved changes independently; preset-only edits no longer mark the RGB effects file dirty
alex — Tools - Test: defer controller connection until a model/port is selected, eliminating startup delay; re-resolve hostnames on demand for controllers that were offline at startup
daryl — Recall the last directory used in Open Sequence and Save Sequence As dialogs (#3107)
Bug Fixes
daryl — Undo of a Model Set move now restores every member of the Set, not just the grabbed prop (#6681)
daryl — Mark the layout as needing a save when aliases are added
charlie — Fix app hang when changing a color to a gradient/color curve on an effect linked to an Effect Symbol (deadlock in symbol propagation)
daryl — Fix Align To Closest Timing Mark doing nothing when both ends of an effect are closest to the same mark (#6659)
daryl — Fix Align To Closest Timing Mark creating impossible effect overlaps (#6652)
cybercop23 — Fix crash configuring a 3D Servo model when servo count is increased (#6663)
cybercop23 — Fix "Copy Layers/SubModels to Models" for submodels with trailing empty layers (#6647)
dkulp — Fix crash cancelling or failing a model download/import with a malformed .xmodel (double-free)
dkulp — Fix crash in 3D layout preview after deleting a model (stale cached model list)
dkulp — Fix crash closing the Key Bindings editor with an uncommitted property edit
dkulp — Fix crash clicking "Render All" while a sequence is closing
dkulp — macOS: fix crash after deleting a model while a property-grid combo popup is being torn down (wxWidgets fork fix)
dkulp — Metal: fix crash when the GPU refuses a texture allocation (memory pressure or zero-sized image)
dkulp — Windows: fix crash rendering Shader effects after a GPU device reset (WDDM TDR)
dkulp — Linux: fix random startup crash from Moving Head panel adding its Color page to a notebook twice (use-after-free)
dkulp — Fix headless render not honoring the "Use GPU For Rendering" preference
dkulp — Fix Galaxy effect rendering extremely slowly on very wide/short buffers (e.g. "Single Line" render style)
Changes
dkulp — Render jobs now suspend and reschedule instead of blocking threads, reducing the thread pool to roughly CPU+GPU core count and lowering memory use on large sequences