# Animation 2025 ## Animation Data Model: Exploration 2 ======== ## Who is this for? --- <p style="text-align: left">On a <strong>data level</strong>, "traditional" 3D animators are already covered:*</p> * Got key frames? Done. * Even actions in their current form are overkill. * Enhancing existing workflows is (mostly) about tooling.* <p style="font-size: 0.4em; text-align: left">* Things like a take system could still benefit traditional workflows. And if you're trying to fix someone else's animation, tossing an animation layer over the top might be appealing even to die-hard traditional 3D animators.</p> --- This is for: * Animators who want **non-traditional** workflows. * **Game animators** building animation libraries. * **Tight deadlines**, combining existing animations. * **Mocap animators** working with, adjusting, and mixing mocap data. * **FX artists** managing baked rigid body sims. * Etc. ======== ## What do we want out of it? --- ### Good handling of multi-ID animation data: * E.g. multi-character animation, animation that includes materials, baked rigid body sims, etc. * In animation libraries. * In NLA workflows. * Pass the "what if bones were objects" test. --- ### Animation layers: * Build up animations in layers. * Adjust existing animations with layers. * Bake down layers and extract data into layers. --- ### Fluid key-frame + NLA workflow: * Should feel like an integrated whole, not two separate systems. * Use manual key frames and assembled clips together. * Bake clips down to key frames and extract key frames into clips. --- ### Takes: * Try out and switch between multiple options. * Keep around the blocking pass for reference. * Not just for the whole animation, but also parts (both time and space). * Etc. --- ### Remapping/retargeting animation data: * Motion capture data. * Updating animation to an updated/fixed rig. * Adapting animation made for one character to another. --- ### Pipeline compatibility: * Animating linked data. * Linking animation into lighting files. * Etc. --- ### Play nice with control rig, dynamic constraints, etc.: * Switching constraints/rigs mid-animation. * Mixing animations from different control rigs in an NLA workflow. --- ### Play nice with Grease Pencil, baked sims, etc.: * 2D-animated characters interacting with 3D props/environments/etc. * 2D FX interacting with 3D characters/props/etc. * Simulate a single dust puff, bake it, and use it for all foot steps (saving simulation time/memory). * Character sitting and smoking a pipe: the smoke puff is part of the animation. Linking the animated character to multiple scenes as background element. --- ### A hatch for future expansion. We don't know what the next 10+ years will look like. Prepare a spot for future data model additions. --- ### Don't lose traditional workflows: * Traditional workflows should remain just as straightforward and convenient as now. (Maybe even get more convenient!) * Any additional complexity in the system should be ignorable. Only pay for what you use. ======== ## Part 1: Slotted Actions ---  Actions have multiple slots. ID blocks point to action **and** slot. --- ### Pros * Stores related key-frame animation data together. * ID-level animation remapping (like current actions). * Pipeline workflows remain straightforward (e.g. linking animation into lighting files). --- ### Cons * Doesn't address animation layers (IDs map to a single slot). * If NLA is still per-ID, doesn't address NLA issues with multi-ID animation. * On its own, doesn't unify key-frame and NLA workflows. * ... ======== ## Part 2: Slotted Actions + Layers ---  Slots are still action-wide. A layer contains animation for one or more slots. --- ### Pros * All the pros of Slotted Actions. * Animation layers. --- ### Cons * More complexity pushed into Actions. * Still doesn't address multi-ID NLA issues. * Still doesn't unify key-frame and NLA workflows. * ... ======== ## Part 3: Slotted Actions + Layers + Clip Tracks ---  "A bunch of tracks, some with keys, some with clips." ---  <p style="font-size: 0.6em">(Not a real UI proposal. Just to illustrate.)</p> --- ### Pros * Unifies animation layers and NLA. * Transparent mixing of key-frame and clip workflows. * Addresses multi-ID NLA issues. * Obvious place for bakes to go (key frame tracks). * Complex built-up animations can be referenced and used like anything else. --- ### Cons * Even more complexity pushed into Actions (but removed from elsewhere). * Extracting keyframes into new clips necessarily creates new actions (data clutter). * Owned clips? * Still doesn't address remapping of animation data on a more-granular-than-ID level. ======== ## Remaining Unknowns --- #### Where do keys go? * When there's no action assigned to the ID block. * One main action? * Create a new action? * When there's more than one track. * An "active" track? What about clip tracks? * How does this affect new-user experience? --- #### What do takes look like? How powerful do they need to be? What will animators use them for in practice? Action-level? Track-level? Clip-level? --- #### How does this integrate with Grease Pencil? Does it need to? --- #### How does this integrate with baked simulations? (Fluid, hair, cloth, etc.) Does it need to? --- #### Tighter integration with constraints / control rigs? Is that needed for the kinds of powerful constraint and rig switching/mixing we ideally want? --- #### How should animation remapping work on a UX level? * An action with 10, 20, or 100 slots. * An action that involves non-object IDs. (Problem not unique to slotted actions.) --- #### How should animation retargeting work? (As distinct from remapping.) --- #### Motion graphics? Haven't sufficiently considered this use case yet. --- #### Bike shed: naming. Maybe "Action" isn't a good name for this anymore. "Animation Sequence"? "Animation Block"? ======== ## Definitely Not Done Yet! Still some unknowns. Still needs more iteration. But hopefully a good direction. ======== ## Thank you!