# Layered Animation workshop 2025 *This is an automatically generated summary of the notes we took on the day. We never got around to writing a blog post about this.* ## The People - Christoph - Demeter (but happy to hand his seat to an animator) - Dorothee - Eos - Falk - Joseph - Nacho - Nathan - Rik (as Blender Studio representative) - Sybren ## What was discussed Two groups tackled the layered-animation UI problem from different angles: - **Group 1** — navigating/editing files that already have messy layers (unnamed layers, multiple actions animating different data unclearly) - **Group 2** — the from-scratch workflow: creating your first layer, adding a second, naming, action/slot assignment Running thread through both: where do layers live in the UI (dedicated editor vs. sidebar vs. a Dope Sheet mode), and how you reveal detail (zoom vs. collapse vs. separate editors). ## Conclusions (+ rationale) - **No new dedicated top-level editor.** Rik's point drove this: animating is a performance activity, and zooming/context-switching between editors breaks focus. Sybren explicitly adopted this as the reason to avoid a new editor. - **Width-responsive layer editor**: below a size threshold it's a flat layer list; above it, a VSE-style layer+strip editor. This sidesteps deciding which editor(s) a small layer manager belongs in — user docks it wherever. - **Dope Sheet/Graph Editor channel list** gets collapsible layer groups (containing the bone channels), plus multi-select layer-solo, mirroring how bone collections already work. - **Global/Active filter modes**, same pattern as today's Dopesheet/Action Editor toggle — Active = current object's action, Global = filterable view across all actions in the file. - **Keys always go to the active object's action**, created if it doesn't exist yet — resolved via Sybren's flow diagram covering 2- and 3-armature edge cases with mixed action states. Unfortunately the diagram itself seems to be lost. - **New actions are just named "Action"**, not derived from the object name — Nacho's reasoning: less confusing. - **A single layer isn't shown by default** to avoid clutter, but must be reachable by expanding/changing the view. - **Editing data outside of strips stays possible** (Christoph) — strips shouldn't gate access to the underlying keys. - **Combined bone transform is viewable in the Graph Editor but not editable for MVP** — unclear semantics for what editing a summed curve would actually write. - **Influence values surface as a normal channel** in the Dope Sheet so they retime along with everything else. - **Keyframe strip type shouldn't disturb pose vs. straight-ahead workflows** — "only pay for what you use." ## Left open / unresolved - Where the "which layer am I on" indicator lives — outliner, side panel, or dedicated UI (Eos raised it, no decision landed). - Copy/paste semantics for layers vs. strips — moving slots is hard since they're strip-bound; might end up layer-only or strip-only copy. Deferred. - Whether layer naming should affect the Merge Animation operator, or just fall back to `Name.001` collisions. Deferred. - Whether control-to-layer binding via 3D viewport selection is *the* design or an exceptional case — no resolution recorded. ## Action / attention points - **Nacho: ship behind an experimental flag immediately**, no exceptions. - **Explicitly cut from MVP**: editable combined-transform curves, layer color/pattern coding, per-bone "show all contributing layers" mode, VSE-style key retiming operator. - **Hard requirement, not nice-to-have**: both Dope Sheet and Graph Editor must display/edit across multiple actions+layers simultaneously — an editor limited to one layer/action at a time is a no-go, so filtering has to be genuinely powerful. - **Baking is a hard mocap dependency**, and flagged as getting more complex once non-keyframe strips exist. - Take the wireframes + group conclusions into the conference presentation (per the day's schedule).