Choose animations for your game

For a third-person action prototype with controller-driven movement, inspect Protofactor Basic first for directional walk/run/crouch. For a smaller walk/run prototype or rifle running, inspect the corresponding Mixamo set. These are starting points based on the evaluated content, not rankings of animation quality. Neither collection has an accepted complete gameplay controller.

Use the comparison below to shortlist content, then open its report for the actual clips, known problems, and setup. The collection catalog is the complete list of evaluated packs. For custom combat timing or a distinct style, compare a library plus authored changes against authoring in Cascadeur before committing to either approach.

Compare the features you need

This example brief is a humanoid action game with a kinematic controller (game code owns movement and collision), directional locomotion, melee or ranged combat, aiming, dodges, jumps, and reactions. Select the features your game needs; a focused pack is not defective for omitting others.

Candidate means content worth testing, not accepted gameplay. Named only means filenames or inventory groups suggest a use that has not been validated. Unevaluated means the reports cannot answer, not that the product lacks it.

Camera: these evaluations concern full-body humanoids for visible characters. Appearance through a third-person gameplay camera is untested. Dedicated first-person arms/viewmodel content and suitability are unknown; full-body motion may still be relevant to NPCs or a visible player body in a first-person game.

Scope: Ultimate advertises 24 animsets in the dated listing, with eight locally evaluated here. Mixamo's nine local pools are not a verified vendor bundle or exhaustive catalog. Follow the collection reports for exact scope.

Use a library, customize it, or author in Cascadeur

Choose the route using one representative gameplay loop on your own character:

• Library first: useful when its selected movements match the prototype. Inspect the original clips, configure movement/looping, and test one coherent set. Count retargeting, contact and transition work before expanding to more weapons or states.
• Library plus authored changes: worth trying when the base movement fits but a specific transition, grip, attack timing or style needs work. First distinguish unclassified content from a genuinely missing motion. Keep useful source clips and author only the demonstrated gaps.
• Author the motion set: worth comparing when required motion or style cannot be achieved economically from the tested library candidates, or when exact gameplay timing is central to the design. You must create and validate the directional variants, transitions, contacts and action recoveries as well as the headline walk or attack.

Cascadeur documents character animation authoring, animation import, retargeting between rigs, and export. These official pages were checked on 2026-09-23. They establish possible workflows, not a tested round trip for these packs or your character. No Cascadeur authoring, editing, export or effort trial was performed for these reports. Confirm the needed features in your edition before choosing that workflow.

Which costs less work? Unknown. There is no common-character, common-engine comparison of library integration, library customization and original authoring. AnimSmith can check declared motion and produce specific mechanical candidates; it does not choose artistic intent or author missing motion. A clean source check does not establish good gameplay motion.

What can I use now, and who must fix it?

Each report's Adoption decision describes one set and the remaining work:

• Inspect unchanged source: a diagnostic playback experiment is possible; this is not approval to ship or to include the set in the finished controller.
• Configure the game: choose movement ownership, looping, blend parameters and state transitions. This changes interpretation, not source animation.
• Try an AnimSmith candidate: a named operation improved a stated measurement. Check the resulting motion, contacts and transitions before replacing the source. An output file alone is not a successful fix.
• Repair or replace source: a confirmed defect requires the specified artist or vendor action. For example, Shield's root-motion crouch set contains a two-bone member; that blocks that set, not every Shield animation.
• Resolve missing evidence: inspect clip intent or run the named test. An untested layer, endpoint hold or phase difference is not automatically an artist defect.

The issue register records severity before changes. A demonstrated workaround may reduce the remaining impact, but only after the relevant checks pass. An unaccepted candidate leaves the outcome unresolved. Artists should read the exact affected clips, current action and acceptance condition before editing.

Do not rank libraries by warning totals or headline verdicts. The reports use different declared expectations and have different engine-test coverage. “Restricted use” and “Usable with conditions” describe scoped decisions, not comparable quality scores.

Run a small comparison before committing

Use the same character, camera, engine, controller speeds and acceptance criteria for each route. Start with idle → move/turn → one action → return to movement. If aiming while moving is essential, include it in this first experiment rather than deferring it until after buying or authoring the full set.

1. Record exact clips and which motion is absent, unclassified or unsuitable. Begin with the report's linked members. A missing report entry is not proof that a vendor lacks that motion.
2. Assign translation, yaw and collision to one owner per state. For a kinematic prototype, start with controller-owned travel and pose animation; test any animation-driven action as a separate mode.
3. Test rapid direction/speed changes, starts/stops and interrupted actions. Check foot sliding, pose pops, ground contacts, hand/weapon grip and deformation through the gameplay camera.
4. When testing a cross-pack handoff, hold gait and speed constant where the source supports it. A walk-to-armed-run example changes two variables; isolate each state's behavior first, then test the combined transition.
5. Record preparation, retargeting, source edits, controller setup, authoring and retesting effort separately. Keep failed attempts and unresolved issues in the comparison. Repeat the same gameplay loop for a custom-authored alternative before concluding it saves time.

A useful result is: selected set and character; unchanged/configured/edited source; checks passed; visible problems remaining; next owner; hands-on effort. Set tolerances for your game before comparing results. For a complete intake procedure, continue to the game-developer workflow.

Read a report efficiently

Read Technical decision for the first experiment, Capability coverage for other candidate motions, and Runtime sets and authored motion for each set's adoption path. Use the Technical issue register in the selected report for blockers and repair instructions. Engine status tells you which tests actually ran.

Open Detailed evidence for exact filenames, timings, grouping assumptions, commands and provenance. Historical evaluator results remain under Changes between AnimSmith versions. This guide and the report prose were refreshed on 2026-09-23 using retained evidence; the measurement dates in each report remain the dates of those tests.

In the reports, a hypothesis is a proposed grouping to test; a declared contract is the movement, timing or loop behavior a check was asked to enforce. A passed contract is limited to those checks. The readiness ladder explains the steps from readable files to accepted gameplay.