Playable build test report
All 26 packaged-build checks passed, with no console or asset errors. The separate controller/camera/audio suite passed 13 checks.
Hardware: Ryzel, AMD Ryzen 9 9900X3D, NVIDIA GeForce RTX 5080 (D3D11), Windows 11 Home 64-bit, approximately 64 GB installed memory. Frame timing is measured from presented requestAnimationFrame intervals, under display synchronization; it does not establish uncapped GPU headroom.
Measured performance
| Preset | Render resolution | Window | Mean FPS | 95th percentile frame |
|---|---|---|---|---|
| low | 1267 × 736 | 7.2 s | 60.00 | 16.80 ms |
| balanced | 1584 × 921 | 7.2 s | 59.99 | 16.80 ms |
| high | 2376 × 1381 | 7.2 s | 60.00 | 16.80 ms |
| balanced-soak | 1584 × 921 | 60.6 s | 59.99 | 16.80 ms |
| balanced-maximized | 3840 × 2089 | 8.1 s | 59.99 | 16.80 ms |
Functional checks
| PASS | All runtime art/audio load without errors |
| PASS | Three material finishes use the same character rig |
| PASS | Mouse capture and look, camera-relative WASD |
| PASS | Ground jump, double jump, third-jump rejection and landing |
| PASS | Distinct double tap: forward-roll |
| PASS | Distinct double tap: roll-left |
| PASS | Distinct double tap: roll-right |
| PASS | Distinct double tap: backflip |
| PASS | Continue holding W: roll-to-sprint, release to stop sprint |
| PASS | Operating-system held-key repeat does not roll |
| PASS | Short route: ramp, platform, actual four-meter double-jump gap |
| PASS | Roll and sprint stop at solid cover; no wall penetration |
| PASS | Camera collision contracts before wall and recovers in clear space |
| PASS | Weapon: shotgun fires, hits and drops a target |
| PASS | Weapon: machinegun fires, hits and drops a target |
| PASS | Weapon: flamethrower fires, hits and drops a target |
| PASS | Weapon rays are stopped by cover |
| PASS | Pause freezes simulation/audio and releases mouse; resume recaptures |
| PASS | Backspace resets position, held action state, route and targets |
| PASS | Repeated movement/jump/evade/fire transitions leave no stuck state |
| PASS | Two-hand weapon alignment at idle and while firing |
| PASS | Measured low graphics performance |
| PASS | Measured balanced graphics performance |
| PASS | Measured high graphics performance |
| PASS | Sixty-second balanced movement and weapon soak |
| PASS | Default maximized window at Balanced settings |
Character and sound
One shared rig, three finishes, authored compact tuck and source locomotion/jump/landing clips prepared in Blender. The weapon uses two-hand IK, and the controller keeps its collider upright during flips. Contact-window measurements informed the configurable stride rates; this prototype does not include terrain-aware foot IK.
All 107 local audio files decoded. The main banks contain 30 independent action vocal takes, 30 other independent jump/evade takes, and 30 designed cloth/impact variations. The foley files are layered edits of real CC0 recordings. Game audio is included in the gameplay recording.
Remaining limits
- Laptop performance is untested. Low, Balanced and High presets are provided.
- Minor contact/stride imperfections can remain on abrupt steps and direction changes; there is no terrain-aware foot planting.
- The portable Windows build is unsigned. Local launch on Ryzel was tested.
- This is a compact practice arena with resettable targets; broader gameplay systems are outside this proof of concept.
The input tests use actual keyboard/mouse events in the packaged executable. QA hooks place the character for isolated collision/weapon cases and select a camera aim point. The ramp and gap route is traversed using real movement/jump inputs. Screenshots and the recording show the actual build.