REBUILT 2026 render assets — v0.12.4 / Onboard Cameras

Onboard cameras (v0.12.4): two first-person views join the match camera selector. Intake looks over the front roller; Shooter looks over the firing mechanism and follows turret azimuth where fitted. Both use the rendered robot pose, including tilt and replay interpolation. Keys 5 and 6 select them directly; Tab, the Xbox View action and the touchscreen camera control cycle all six views. Keyboard shortcuts and the gamepad camera action remain remappable. Switching views preserves robot-relative driving. Onboard mounts are virtual game placements, not claims about team camera hardware. The same single render pass is used; no video feed or second scene renderer is added.

Fuel cycle (v0.12.3): the opening stages 360 neutral-zone FUEL in the manual setup envelope. Robots collect individual balls through their deployed intakes into storage. Every launched ball continues below the HUB rim, through the internal return path, out of one of four neutral-facing exits, then bounces and rolls on the carpet. The scoring camera moves around to show the outlets. Launched identities persist until the scene cut; each clip reserves ammunition within the existing 504-instance pool. Pausing and seeking remain deterministic. Exit positions are aligned to the field CAD; return timing, spread and bounce are original cinematic approximations. Match physics, AI and scoring are unchanged.

2026 Game Manual / 比赛手册 §§5.4, 6.3.4–6.3.4.1 · Reviewed / 核对日期：2026-09-09.

Continuous feed (v0.12.2): 1678 and 8214 now emit individual FUEL instead of synchronized three-ball packets. 1678 uses a dense flow; 8214 has stronger short-term changes in spacing. Outlet positions vary across the available width and flywheel speed varies smoothly under load. Assisted shots re-solve their trajectory to the same HUB. Nominal game rates remain 24 and 21 ball/s respectively; these are authored simulation settings, not measured team telemetry. The opening uses the same feed profiles with separately phased internal balls. Other robots retain their metered or parallel match delivery. Releasing fire, an empty hopper and aiming interlocks cannot accumulate a catch-up burst. The full FUEL pool and one-pass opening renderer are retained.

Feed references reviewed 2026-09-09: 1678’s official 2026 robot description documents its full-width shooter, roller floor and accelerating ball tunnel; 8214’s team CAD release provides structural references. Neither source is treated as a measured firing-time trace. The different timing profiles are game interpretations of the user-requested continuous delivery.

Scoring cinema (v0.12.1): the return sequence brings 4414 from the neutral zone into its own alliance zone, then shows internal feeding, shooter spin-up and a volley into its own HUB. Every opening shot is launched from the visible outlet with a fixed origin, aligned hood angle and a continuous descent inside the rim. Closer cameras, restrained fill lighting, visible feed balls, wheel rims and shaft detail make the shooting mechanisms easier to read. The opening uses an authored demonstration cadence; it does not change match throughput, physics or scoring. Models, trajectories and choreography remain original game approximations, not exported team CAD or a recording of a real match. Browser checks cover all eight robot outlets and feed reset at both quality levels, plus desktop and portrait cinematic framing. Physical tablet frame rates have not been measured.

Robot silhouettes (v0.12.0): rebuilt storage proportions, front retaining walls, rear shooter bays and full-height feeder supports. 1678 has a white expanding hopper and a broad rear shooter; 8214 has amber walls, an arched net and exposed spoked flywheels; 7769 has a lightened metal bin and a wide blue wheel bank; 4414 has a rear portal, annular feed and elevated turret. 9483, 581, 2910 and the Everybot reference retain distinct feeding layouts. Near views add cut sheet plates, hollow frame tubes, bearings, motors and drive belts. The 1678 public Onshape assembly was inspected interactively on 2026-09-09; other models use the CAD renders and photographs listed in the model reference page. No team CAD meshes or source photographs are redistributed. Dimensions and motion remain game approximations; a 60% fidelity score has not been measured. The shot origin, F lock and moving collision parts now share the revised layout. Fixed shooters fire from the rear; the intake remains the driving front. Low quality renders 5,080–6,312 triangles per robot; high quality renders 19,768–29,044. Distant service hardware is hidden. Software-renderer checks cover all eight robots, three viewing angles and shot outlets; these are not physical tablet frame-rate measurements.

Character detail (v0.11.3): spectators, drive teams and referees now use cropped anatomical head meshes, shaped clothing, separate fingers in near views, safety-glasses frames and more varied body poses. The head and scalp geometry is adapted from MakeHuman Community CC0 graphical assets; clothing, props and poses are original game work. All characters are fictional, not portraits. High quality reserves richer heads and hands for crew, officials and the front row; background characters use simplified geometry. People, stands and desks total 13 / 10 instanced draws and 463,240 / 178,248 triangles in high / low quality. They add no physical colliders or per-person shadow passes. Browser checks cover close views and all six driver positions; physical iPad / Android frame rates are not yet measured.

MakeHuman graphical data: pinned asset license, sections C–D; CC0 notice. Source commit a8bc2d54ff0ac92e78ff71431b1023eda42bf482; base.obj SHA-256 8e761e6624b8f54536409135d1636da63b32486a90d4897f84e121d144f6fb4c. Only cropped head and scalp vertices are redistributed in people-anatomy.mjs; no full body, MakeHuman application code, or real-person photographs are bundled. Extraction, decimation, preserved surface normals and coordinate conversion: scripts/prepare-people.mjs. Retrieved 2026-09-09.

Match moments (v0.11.2): the opening starts inside the 2026 field. Four seven-second scenes show neutral-zone collection, staggered scoring volleys, a neutral-zone feed into an alliance intake, and a late tower lift while teammates finish shooting. The finale selects one climb-capable game robot per alliance; all-shooter rosters keep shooting. Lift straps are original cosmetic game rigs, not team CAD. Cameras, ball trajectories and timing are authored for this introduction; this is not an exact reconstruction of a historical match. Pause, Next scene and immediate menu entry remain available. Reduced motion uses a static field view. Reused models and fixed pools keep the title to one scene render per frame. The montage never advances match physics, scoring, AI or saved sessions.

Inspiration: FUN Robotics Network’s March 26, 2026 publisher post describes 1678 / 5940 / 199 alliance coordination in the Sacramento high-score match. The post was reviewed; video frames were unavailable for frame-by-frame verification. No video or real result is reproduced. The 2026 Game Manual, sections 5.8, 6.5.2 and G407 informs tower staging and alliance-zone shooting.

Retrieved 2026-09-06. This is an independent training game; no endorsement is implied.

Presentation update, 2026-09-08: The title now opens with a bottom-centered Chinese / English / Turkish prompt. Click, tap, Enter, Space or gamepad A opens a continuous robot-centering transition, followed by Play, Choose Your Robot and Settings. Reduced motion uses a static prompt in the preferred interface language and an immediate reveal. Mode selection retains the in-game field camera; the robot menu renders the selected game robot. It is not match photography, season-history footage or a Canva-generated image. Creative mode uses original ability names, mechanics and bounded Three.js effects. The official RoboMaster 3v3 overview informed the general combat-mode direction; its artwork, robot models and official skill rules are not bundled. Classic FRC rules remain separate. Original contributions: © 2026 Harry Xin / 6941 IronPulse.

No runtime asset fetches leave this site. Models, tags, libraries and license notices are included in the static distribution.

Event Floor model rebuild

Eight upper assemblies were rebuilt following the per-model evidence audit. See ROBOT-MODELS.md for current inspected references, visual changes and adaptation limits. The older AUTO update notes below describe the previous implementation.

Creator, robot references and presentation

September 7, 2026 robot and AUTO update

© 2026 Harry Xin / 6941 IronPulse, for original project contributions to the extent copyright applies. AI-assisted development. Third-party rights and licenses remain separate. See the bilingual copyright and sources page.

Event people and officials (updated in v0.11.3)

The 2026 Field Manual §3.5.2, figure 3-21 (page 40) identifies five referee panel posts: HR, RN, RF, BN and BF. Game Manual §5.13 places the yellow-shirted head referee between the field and scoring table. This version adds one head referee, four side referees, their five panels and a scoring desk. Exact placements are adapted to this game field and keep the driver stations clear. Original fictional characters have tapered clothing, articulated limbs, varied hair and poses, and additional front-row facial detail. Spectators remain 132 / 64 in high / low quality, plus 30 drive-team members and 5 officials. All people, stands and desks use 13 / 10 instanced draws (463,240 / 178,248 triangles); they have no physical colliders or per-person shadow passes. Driver and head-referee bodies are hidden at their own first-person camera. These are cosmetic staff models, not additional automated officiating rules.

References: 2026 Field Manual, p. 40 and Game Manual §5.13, §§6.2–6.3. This earlier procedural audience design has since been updated with the CC0 human bases credited below. Official clothing logos are not bundled.

Film presentation and match announcement — September 12, 2026

The friends-v2 trailer and promo-v4 introduction use original 3D DL and DriverLab brand geometry, beveled metal surfaces, studio reflections and shadows. No DJI or FIRST logo, commercial film footage or announcer recording is included. Generic Yunyang and Andrew synthetic voices provide Chinese and English narration; the music is the original score Drive the Alliance. Production: Harry Xin / 6941 IronPulse. Assistance: GPT-6 Astra.

The multiplayer trailer records the actual game interface and two local Colyseus clients on a synchronized offline 60 Hz timeline. It is not a hardware performance benchmark. The climb engagement position is staged before the cut; the lift and scoring use game physics. The introduction retains the previously approved match shots. Both films provide full spoken captions and editable SRT, VTT and TXT files. Copyright and model-source credit pages last two seconds each.

The English Drivers behind the line call uses an original synthetic voice. Its countdown display follows the player's English, Simplified Chinese or Turkish setting. Start, AUTO-end and match-end tones are original procedural sounds.

Arena interiors and Combat v3 — September 13, 2026

Shanghai is an original geometry interpretation of the white gym walls, high windows, roof trusses, equipment and crew circulation visible in the SMG / CCTV Shanghai report. The official Shanghai event page identifies the ECNU Minhang venue. The official Istanbul Final 2 video public preview informs dark upper seating and pale stair aisles; only the preview frame was inspected for that venue. The FIRST Einstein seating plan informs stage, crew desk and divided seating. Other halls remain generic interpretations. These are not surveyed replicas. Reference footage, broadcast overlays and logos are not included as scenery.

Heads now share their neck geometry and transform. Existing MakeHuman Community CC0 head-base attribution remains applicable; the clothing, bodies and venue structures are original game geometry. People use 465,768 / 180,030 triangles in high / low quality, with the existing instanced draw budget and no person colliders. Hall boxes and lights use two bounded instanced draws.

Combat v3 is fictional: six minutes, first to 30 robot defeats, 300 initial robot HP, 900 Hub HP, five-second respawn. A breached Hub permanently reduces its alliance's robot health cap to 150. Equal defeat totals at timeout or a same-frame double threshold draw. Artillery uses 120 direct damage, a 1.6 m falloff radius, eight ammunition and a 20-second cooldown. These are game balance values, not real robot capabilities or official FRC rules.

The 108-second bilingual main film uses current game physics, staged initial positions, current three-client room UI and offline 60 Hz frame capture. Phone/tablet input is emulated. Original published narration is retained; four new bilingual Combat sentences use generic synthetic voices, with matched loudness and an original music bed. No DJI voice recording or cloned actor is used. Complete editable captions and two-second credit pages accompany the film.

Presentation and six-film revision — September 13, 2026 (v0.20.2)

Overhead cameras cut away camera-facing venue walls, roof trusses and ceiling fixtures. These are venue structures, not field tape or game objects. Driver views retain the hall. Hub lights reuse eight diffuser surfaces extracted from the existing licensed field CAD and simplified to about 1,724 triangles. The WPILib / AdvantageScope BSD-3-Clause attribution above also covers these surfaces. No official logo or broadcast overlay is added.

Classic lighting follows 2026 Game Manual TU22, §5.4, Table 5-3: alliance color when active, warning pulses before deactivation, transition white chase and white during final scoring assessment. This is availability information, not a per-ball flash. Combat uses a separate HP bar and darkens a disabled Hub.

The arena-v022 edition replaces all three Chinese and English films: promotion (71.5 s), rules (102.5 s), and controls (87.5 s). Narration and full captions are rewritten; generic Yunyang / Andrew synthetic voices are mixed against the original Drive the Alliance score. No actor is cloned. Creation: Harry Xin / 6941 IronPulse; assistance: GPT-6 Astra. Each credit page lasts two seconds.

World footage is native 3840 × 2160 at an offline 60 Hz simulation/render step. The room and replay use three local clients and three AI seats. Touch capture uses browser emulation and is scaled into the 4K master. Menu selection is held for reading. Initial fuel, hopper fill, separated climb positions and near-breach Hub health are staged; subsequent actions use game physics. This is not a physical-device performance measurement. Intake and climb shots are checked for rail/robot contacts; the AUTO demo assigns a stationary center preload with separate left/right sweeps and records zero teammate contacts.

The broadcast camera is centered on the field axis, keeping long field edges level in live play, replay and the replacement footage. The logo sequence has a dimmed overhead arena background. Documentary-style credits use two compact, left-aligned pages. Chinese captions use Noto Sans SC Medium; English uses Noto Sans, with a fine shadow. Fonts are licensed under the SIL Open Font License. This is an original layout inspired by contemporary film typography, not an assertion that these are another creator's exact fonts or title designs.

The multiplayer lobby uses larger individual camera crops in one on-demand renderer; phone layouts use two alliance columns. The host's Start confirms their readiness, while the server continues to validate all participants.