Animation pack evaluation: Protofactor Ultimate Animation Collection (partial: eight packs)
Technical verdict: Insufficient technical evidence
Evaluation completeness: partial — source and declared checks were rerun for eight animsets with AnimSmith 0.14.0; proposed cross-pack states and visual/gameplay behavior remain untested.
Confidence: medium
Evaluation date: 2026-09-16
Current evaluator: AnimSmith 0.14.0
Report format: 3
Detailed evidence: partial collection evidence appendix
Technical decision
Prototype decision: For a controller-driven (kinematic) pilot, start with unchanged Basic in-place walk/run sources on the target character, then add one full-body melee state. The first test is direction and speed sweeps plus the armed idle/walk handoff, checking foot slide, grip, pose pop and collision. Resolve intended loops and contacts before controller admission. Measured AnimSmith outputs are optional trials that still need those checks; source correction is required where a confirmed seam or malformed hierarchy blocks the selected set. In-place clips leave travel to the controller; root-motion clips carry authored travel.
Camera scope: Full-body humanoid animations; appearance on your target character and camera is untested. Dedicated first-person arms/viewmodel use is unverified.
As of the 2026-09-17 vendor listing, Ultimate advertises 24 animsets. Eight constituents were evaluated here; 16 were not: Combat Bare Fists, Fencing, Wizard, Bow & Arrow, 2 Handed Gun, Assault Rifle, Bazooka, Dual Guns, Minigun, Shotgun, Hostage, Zombie, Creature, Crowd, Push & Pull Cube, and Female Basic Locomotion. The older local report named “Double Guns”; its relationship to the listed “Dual Guns” is unverified. Current listing membership does not establish that the unknown local revision contained all 24.
Content in evaluated inventory: Eight evaluated constituent state families. Ground locomotion, injured styles, three melee styles, shield combat, climbing and campfire interactions have selected candidates; sixteen currently advertised constituents are unevaluated. These are identified motions; gameplay behavior has not been approved.
The official AnimSmith 0.14.0 release reran Basic Locomotion, Sword & Shield, Campfire, Climbing, Injured, 1-Handed Melee, 2-Handed Melee, and Dual Swords. They contain 918 source candidates, including 895 individual motion-labelled inputs. This is current mechanical intake evidence, not collection-level approval.
Can these packs form a kinematic animation controller? They provide useful candidate ingredients, but the evaluated combination is not an accepted controller. Use Basic Locomotion as the ground state, Injured as separate style states, melee packs as full-body armed states, and Climbing/Campfire as constrained traversal/interaction states. Shared clip naming and successful import cannot establish smooth handoffs, matching contacts, or compatible upper-body layers.
The constituent reports now name exact current scenario members and actions. They do not reconstruct a missing historical collection manifest. All 159 transform trials still need the relevant in-game checks before any output can replace its source.
Capability coverage
Content present
Eight locally evaluated constituents offer selected ground gait rings, seven injury styles, four combat families, wall/ladder traversal and campfire interaction chains. Their classifications are evaluator hypotheses tied to constituent source inventories.
Content gaps and unknowns
Sixteen animsets on the 2026-09-17 vendor listing were not locally evaluated, including Female Basic Locomotion. Their content, overlap and compatibility are unknown for this local revision. Missing evidence is not missing product content.
Evaluation still needed
Test one cross-pack state boundary at a time, then target-character blend, contact, root ownership, retarget, visual and build behavior. The partial source and Unity probes do not establish collection-wide controller acceptance.
Action-game inspection leads
These retained groupings guide source inspection, not controller acceptance. The selected gait recipes below have exact members; the action-group rows generally do not publish member-level bindings or interruption contracts.
| Need | Source lead and current status | Next check |
|---|---|---|
| Sprint and turns | Basic filenames include sprint, fast-run, U-turn and 90/180-degree turns outside the selected rings; candidate, untested | Bind takes, direction and root ownership; test stops/turns under player input |
| Dodge and parry | Two-Handed dodge-forward-back and parry-3-way; candidate groups, untested | Bind exact members, windows and contacts; check interruption and collision |
| Attacks and combos | Dual Swords combo-alternatives and single-attack-alternatives; candidate groups, untested | Bind attacks, timing, grip, hit events and cancel rules |
| Equip and recovery | Two-Handed, Dual Swords, and Sword & Shield hold draw/put-away groups; Sword & Shield also groups four death/downed/recovery chains. Candidate groups, untested | Bind source takes and test whole entry, interrupt and exit paths |
Selected gaits and candidate actions answer different questions. A declaration pass or Unity source sample permits a diagnostic trial; controller admission needs intended contracts, target-character visuals, contacts and the actual state graph. For library, hybrid and custom-authoring choices across Protofactor and Mixamo, use the shared evaluation guide. Comparable effort has not been measured.
Runtime sets and authored motion
Use each row as a separate blend or sequence proposal. The linked appendix names its exact clips, timings and measurements; controller playback and contact quality still need testing.
| Set | Controller use | Adoption decision | Exact members |
|---|---|---|---|
| Basic locomotion | Ground walk/run/crouch; controller travel | Start a controller-owned source prototype; phase-alignment output still needs in-game checks and 22/24 candidate outputs retain lint across its three rings. Set loop/contact and root policy. | Constituent exact members |
| Injured | Seven style-local idle/walk/run states | Add one style-local source state at a time; 20/21 selected files fail declared contracts. Review true loops and healthy-state handoff. | Constituent exact members |
| Sword & Shield | Full-body shield combat; in-place or root-motion gait | Choose in-place or root-motion source gait; no named set passes its declared checks. The root-motion crouch set is blocked by a two-bone source member; review one-shot policy separately. | Constituent exact members |
| 1-Handed Melee | Full-body one-hand combat; in-place or root-motion gait | Choose one source gait variant; no named set passes its declared checks. Review continuous seams and isolate the 73-bone blocked state for import/retarget. | Constituent exact members |
| 2-Handed Melee | Full-body two-hand combat; in-place or root-motion gait | Choose one source gait variant; no named set passes its declared checks. Review continuous seams and isolate the two 56-bone block states. | Constituent exact members |
| Dual Swords | Full-body dual-weapon combat; in-place or root-motion gait | Choose one source gait variant; no named set passes its declared checks. Review continuous seams, dual-grip contacts and full-body handoff. | Constituent exact members |
| Climbing | Separate wall and ladder traversal | Prototype wall and ladder separately. All ten selected members fail declared-loop contracts; pruning did not clear the WallClimbUp loop findings. Geometry/contact remains open. | Constituent exact members |
| Campfire | Posture and prop interaction chains | Prototype discrete interaction chains. Configure anchors/props/exits; kneel hold has a rotational seam that pruning did not clear. | Constituent exact members |
Integration recipe
- Members/topology: Keep ground locomotion, injury styles, armed stances, traversal and interactions as explicit states. Use direction/speed trees only within a reviewed compatible set.
- Timing/synchronization: Confirm intended loop/contact correspondence before choosing normalized-phase synchronization, then test outgoing/incoming poses for state handoffs. Do not equate identical durations with matching support feet.
- State ownership: Select in-place ground motions for a kinematic capsule. Declare any root-motion action as a separate movement mode, including collision and return of authority.
- Composition constraints: Armed/unarmed full-body handoffs are the starting hypothesis. A weapon mask needs explicit spine/root ownership, grip, stance and foot-contact checks before adoption.
- Acceptance gate: Run speed/direction sweeps, starts/stops, interruptions, attachment and geometry tests on the intended character and engine build.
Technical issue register
Severity is scoped to the intended combined controller. A missing test is an adoption risk, not evidence that the vendor authored bad motion.
| ID | Severity | Problem and impact | Primary owner | Current action | Future AnimSmith potential | Evidence/status |
|---|---|---|---|---|---|---|
| UC-PHASE | note | Evaluator-selected locomotion groups have measured phase differences; intended shared contact phase is not authoritative, so no artist defect follows from the spread. Readiness guidance. | unknown | Confirm project or vendor synchronization intent, then test the exact constituent sets against declared phase/contact tolerances. Treat per-pack loop findings separately and request artist changes only where intended playback establishes a source defect. Slicing and phase-alignment outputs still need gameplay checks. | Declared anchoring can reduce measured spread; it cannot choose or approve intended contacts. | observed-animsmith measurements; synchronization/contact acceptance not-evaluated. |
| UC-CONTROLLER | major | Arbitrary cross-pack blending has no accepted movement, transition or contact contract. Guidance: not applicable. | engine-config | Build the explicit state topology above and test exact handoffs. Residual unknown until project acceptance; no pack defect inferred. | Explicit collection contracts can record selected relationships and diagnostics. | inferred integration guidance; controller acceptance not-evaluated. |
| UC-MASK | major | Full-body weapon/interaction motions may conflict with locomotion when masked at the torso. Guidance: not applicable. | engine-config | Start with full-body state changes. Check any upper-body mask against grip, stance and planted feet before using it; compatibility remains unknown. | Mask diagnostics could expose affected channels; authored style/contact judgment remains. | Masked artistic/contact fit not-evaluated. |
| UC-SCOPE | minor | Sixteen currently advertised constituents are outside this evaluation; whole-collection coverage and value cannot be concluded. Guidance: not applicable. | unknown | Consult the explicit exclusion list and evaluate only needed missing gameplay capabilities. Residual scope limitation. | No tool can infer evidence for missing content. | not-evaluated; boundary, not vendor defect. |
Engine status
| Runtime | Evidence level | Technical result | Remaining gate |
|---|---|---|---|
| Unity 6000.5.8f1 | observed-engine | 26/26 bounded source checks: 13 samples, nine Basic/Sword & Shield-to-melee mixer schedules, four root-motion samples. Injured, Climbing and Campfire handoffs were not tested. Finite poses/fixed owner under explicit policy. | Visual/contact, actual controller, target character, compression, candidate output and build acceptance. |
| Unreal Engine | not-evaluated | No current import or playback run. | Import, retarget, graph, contact, and build tests. |
| Godot | not-evaluated | No current conversion, import, or playback run. | Conversion/import, graph, contact, and export tests. |
| Bevy | not-evaluated | No current conversion, addressability, or runtime run. | Conversion, target mapping, runtime, and performance tests. |
Fit and limitations
| Combination | Practical decision | Remaining acceptance |
|---|---|---|
| Basic + Injured | Candidate healthy/injured state handoff; keep injury letters separate | Speed/phase calibration, stance continuity and interruptions |
| Basic + Sword & Shield / 1-Handed / 2-Handed / Dual Swords | Candidate armed/unarmed full-body state change | Weapon grip, stance, turn/stop continuity; masking is a separate decision |
| Basic + Climbing | Ground-to-traversal state handoff with environment anchor | Entry/top-out, wall distance, contacts and collision ownership |
| Basic + Campfire | Stop at interaction anchor, then enter explicit posture sequence | Seat/fire placement, props, interrupt and return-to-ground behavior |
| Protofactor + Mixamo | No approved cross-library handoff | Target rig, rest pose, scale, root convention, phase, style and attachment tests |
Fresh Unity 6000.5.8f1 source tests passed 26/26 bounded checks: 13 individual Humanoid samples, nine named cross-pack mixer schedules and four root-motion samples. The nine mixers cover Basic idle/walk and Sword & Shield idle paired with 1-Handed Melee, 2-Handed Melee and Dual Swords; they do not cover Injured, Climbing or Campfire handoffs. For the nine mixers, five weights (0, 0.25, 0.5, 0.75, 1) produced finite sampled bone transforms with applyRootMotion=false and a fixed Animator owner. This supports technical feasibility of the named source combinations, not visually smooth handoffs or a finished kinematic controller.
The proposed state architecture remains inferred; this bounded feasibility evidence is not an accepted full-controller verdict. The reports are useful for selecting a prototype and assigning author/project work; they do not establish a universal game-ready controller.
Changes between AnimSmith versions
AnimSmith 0.14.0 — revalidated the eight constituent baselines/contracts and 159 bounded remediation candidates. Added exact current generic scenarios and developer/artist decisions.
AnimSmith 0.10.0 and earlier — superseded mechanical and remediation evidence. Historical collection logical-motion/set totals were not reconstructed or promoted into this run.
Evidence status
Current conclusions are constituent-derived and explicitly scoped. Canonical collection taxonomy is unavailable; selected current scenarios are linked without claiming a recovered collection binding. Transformed outputs still need gameplay and visual checks. The readiness ladder and appendix distinguish static, engine-execution and acceptance evidence.
Sources
- Constituent reports: Basic Locomotion, Sword & Shield, Campfire, Climbing, Injured, 1-Handed Melee, 2-Handed Melee, and Dual Swords.
- Protofactor, Ultimate Animation Collection and End User License Agreement.
- AnimSmith, game-ready clips and CLI reference.
- Unity 6 documentation: Blend Trees, Root Motion, and Animation Layers — engine concepts only, not evidence of pack quality.