Awesome Opus 5.5 Prompts

Make an isometric floating mini-world in Blender, animated with Python, as a seamless loop: a small mine island with a terraced mountain, two tunnels and a railway that loops through the mountain, a pond whose stream falls off the edge as a waterfall, and rock layers with glowing crystals on the cut sides. The workers are little Claude bots: one mines a crystal vein and gets startled by a bat, one runs a crane that dumps crystals into each passing cart, one fishes in the pond, and one rides a cart. Put a wooden "TOKENS" sign over the mine entrance. Sound synced to every action.

Can you create a Minecraft clone?

a giant dragon attacking a medieval castle and its village, built in Three.js.

Create a Goal with a token budget of 200,000. At the end, report the actual token usage, budget utilization, and runtime; if input, cached input, and output token breakdowns are available, estimate the cost in USD using the current model pricing and clearly list the calculation basis. I understand this uses subscription quotas, but we can convert them to API billing

Create “Octo Jelly” - a beautiful, interactive 3D gummy octopus that users can grab, stretch, and squish directly in their browser. Deliver a complete single-file HTML experience using genuine WebGPU.
ART DIRECTION
Make the octopus look like a premium translucent gummy candy: a rounded head, eight curled tentacles, small suction cups, and a cute, understated face.
Use a warm off-white background, soft studio lighting, and a subtle ground shadow. Keep the scene elegant and uncluttered, with the octopus large and centered.
GEOMETRY AND MATERIAL

* Generate all geometry procedurally. No external models or image files.
* Connect all eight tentacles smoothly to the body, without visible gaps or floating parts.
* Add rounded suction cups that stay attached as the tentacles deform.
* Use glossy, translucent jelly with thickness-dependent color absorption, refraction, soft internal light scattering, and delicate rim highlights.
* Thick areas should have richer color; thin tentacle tips should transmit more light.
* Avoid opaque plastic, blown-out highlights, and visible mesh seams.

SOFT-BODY PHYSICS
Use a stable mass-spring or position-based dynamics system with elastic constraints and approximate volume preservation.

* The head should feel soft but substantial.
* Tentacles should be more flexible than the head, especially near their tips.
* Allow users to grab the head or any tentacle at the clicked location.
* Pulling should deform the nearby geometry first, then elastically pull the rest of the body.
* On release, the octopus should wobble and gradually settle into its original shape.
* Tentacles should react independently, with slightly delayed motion.
* Include gravity, floor collisions, friction, and damping.
* Prevent tentacles from passing through the floor.
* Clamp extreme stretching and use fixed simulation steps so strong pulls do not break the model.
* Do not fake softness by scaling or rotating the entire octopus.

INTERACTION

* Left-click or touch the octopus to grab and stretch it.
* Right-drag or drag empty space to gently orbit the camera.
* Support a limited zoom range.
* Keep camera gestures separate from object dragging.
* Add “Give it a nudge,” “Reset,” “Pause,” and “Reset view” buttons.
* Include Firmness and Internal damping sliders.
* Add “¼ speed” and “Show mesh” toggles.
* Provide three color presets: Coral, Lagoon, and Grape. Change the material colors without resetting the simulation.

INTERFACE
Use a minimal editorial layout:

* Top left: small “MATERIAL STUDIES” label.
* Large italic serif heading: “Octo Jelly.”
* Caption: “Eight arms. A little wobble. A very soft creature.”
* Top right: a WebGPU status indicator.
* Right side: a compact “THE SPECIMEN” control panel.
* Bottom left: “Grab a tentacle. Pull gently. Let go.”

Use clean sans-serif text for controls, thin borders, and generous whitespace. Avoid heavy panels or decorative UI effects.
PERFORMANCE AND QUALITY

* Use real WebGPU rendering, not a 2D canvas imitation or prerecorded animation.
* Reuse geometry and buffers; do not rebuild meshes during dragging.
* Keep suction cups, eyes, and other details attached to the deforming body.
* Handle transparency without flickering, disappearing surfaces, or harsh black edges.
* Support desktop and mobile layouts.
* Show a clear fallback message if WebGPU is unavailable.
* Test repeated grabs, strong pulls, releases, floor collisions, palette changes, pause, and reset.

The result should feel like a little living gummy toy - glossy, squishy, expressive, and satisfying to stretch. Deliver the full working HTML, not a mockup or a code fragment.

创建一个 token 预算为 200,000 的 Goal。结束时报告实际消耗 token、预算使用率和运行时间；如果能够获得输入、缓存输入、输出 token 的拆分，则按照当前模型价格估算美元费用，并明确列出计算依据。我知道是订阅额度，但是我们可以换算成api计费

Create “Octo Jelly” - a beautiful, interactive 3D gummy octopus that users can grab, stretch, and squish directly in their browser. Deliver a complete single-file HTML experience using genuine WebGPU.
ART DIRECTION
Make the octopus look like a premium translucent gummy candy: a rounded head, eight curled tentacles, small suction cups, and a cute, understated face.
Use a warm off-white background, soft studio lighting, and a subtle ground shadow. Keep the scene elegant and uncluttered, with the octopus large and centered.
GEOMETRY AND MATERIAL

* Generate all geometry procedurally. No external models or image files.
* Connect all eight tentacles smoothly to the body, without visible gaps or floating parts.
* Add rounded suction cups that stay attached as the tentacles deform.
* Use glossy, translucent jelly with thickness-dependent color absorption, refraction, soft internal light scattering, and delicate rim highlights.
* Thick areas should have richer color; thin tentacle tips should transmit more light.
* Avoid opaque plastic, blown-out highlights, and visible mesh seams.

SOFT-BODY PHYSICS
Use a stable mass-spring or position-based dynamics system with elastic constraints and approximate volume preservation.

* The head should feel soft but substantial.
* Tentacles should be more flexible than the head, especially near their tips.
* Allow users to grab the head or any tentacle at the clicked location.
* Pulling should deform the nearby geometry first, then elastically pull the rest of the body.
* On release, the octopus should wobble and gradually settle into its original shape.
* Tentacles should react independently, with slightly delayed motion.
* Include gravity, floor collisions, friction, and damping.
* Prevent tentacles from passing through the floor.
* Clamp extreme stretching and use fixed simulation steps so strong pulls do not break the model.
* Do not fake softness by scaling or rotating the entire octopus.

INTERACTION

* Left-click or touch the octopus to grab and stretch it.
* Right-drag or drag empty space to gently orbit the camera.
* Support a limited zoom range.
* Keep camera gestures separate from object dragging.
* Add “Give it a nudge,” “Reset,” “Pause,” and “Reset view” buttons.
* Include Firmness and Internal damping sliders.
* Add “¼ speed” and “Show mesh” toggles.
* Provide three color presets: Coral, Lagoon, and Grape. Change the material colors without resetting the simulation.

INTERFACE
Use a minimal editorial layout:

* Top left: small “MATERIAL STUDIES” label.
* Large italic serif heading: “Octo Jelly.”
* Caption: “Eight arms. A little wobble. A very soft creature.”
* Top right: a WebGPU status indicator.
* Right side: a compact “THE SPECIMEN” control panel.
* Bottom left: “Grab a tentacle. Pull gently. Let go.”

Use clean sans-serif text for controls, thin borders, and generous whitespace. Avoid heavy panels or decorative UI effects.
PERFORMANCE AND QUALITY

* Use real WebGPU rendering, not a 2D canvas imitation or prerecorded animation.
* Reuse geometry and buffers; do not rebuild meshes during dragging.
* Keep suction cups, eyes, and other details attached to the deforming body.
* Handle transparency without flickering, disappearing surfaces, or harsh black edges.
* Support desktop and mobile layouts.
* Show a clear fallback message if WebGPU is unavailable.
* Test repeated grabs, strong pulls, releases, floor collisions, palette changes, pause, and reset.

The result should feel like a little living gummy toy - glossy, squishy, expressive, and satisfying to stretch. Deliver the full working HTML, not a mockup or a code fragment.

Build with Opus 5.5 a volcano simulation. Magma chamber 5 km down.

Create "Jelly Press": a single-file interactive 3D toy in HTML (all JS, CSS and WGSL shaders inline, no external assets except Google Fonts). Render with WebGPU; if WebGPU or an adapter is missing, show a clean fallback message instead of a blank page.  CONCEPT Four translucent gummy jellies shaped like fruit slices sit one at a time on the steel bed of a hydraulic press. The player holds a big red button to lower the press. The jelly squashes and spreads, the pressure gauge climbs, and somewhere past half its height it bursts into pieces. After the burst the game does NOT end: the player can grab the pieces, drag them, throw them around, and squash them again.  THE JELLIES (chips at the bottom, keys 1–4) 1. Watermelon wedge (half-disc slab): red flesh with dark teardrop seeds, a pale rind band, a green striped skin. 2. Orange slice (half-disc): orange pulp segments separated by thin white membranes, pale pith, orange peel. 3. Fig half: pink flesh full of small golden seeds, a cream layer, a dark purple skin. 4. Pineapple ring: golden fibrous flesh with radial streaks and a hole in the middle. Each jelly should look like real gummy candy: subsurface scattering, soft translucency, glossy specular highlights, soft shadows on a warm studio floor (cream/beige, tone-mapped).  PHYSICS (CPU, fixed 60 Hz step) - XPBD tetrahedral soft body with 8 substeps: co-rotational per-tet shape matching, per-tet volume constraints, hard edge strain limits (0.35×–1.8×), edge velocity damping, floor contact with Coulomb friction, rolling resistance, and gentle settling when nearly still. - Render mesh skinned on the CPU through barycentric embedding in the tets; normals recomputed from triangles every frame. - The press die is a kinematic round platen (radius ~1.05, rounded edge, a thickness, a ram above it). It acts as a ceiling with friction below, a shelf on top, and a side wall at its rim. The two press posts are solid. - Pressure readout in bar comes from the platen's contact load, scaled per fruit.  THE BURST - Break at a random squash between 52% and 66% of the jelly's height. - Plan the fracture in the background shortly after each round starts, so the burst itself is instant. - 5–7 big pieces from 3D Voronoi cells with slightly tilted walls. On 3–4 of them, a far corner is chipped off by two cutting planes and split further into 2–4 small chips, which leaves jagged, notched edges. - Assign tets to cells by centroid. Duplicate particles per chunk. Merge tiny islands into their neighbours. - The new body adopts the old positions and velocities. - Clip the skin triangles against each cell's half-spaces, and fill every cut face with a clean flat cap showing the fruit's interior (flesh, seeds, membranes). No stretched triangles, no holes. - Kick pieces outward and upward from the press. Small chips fly faster and higher and tumble with random spin. - Show a big italic verdict for about 2.5 s, then fade it: "Splat." (melon), "Squeezed." (orange), "Well, that's jam." (fig), "Crushed." (pineapple). Include a stat line: "Gave up at N bar and N% of its height."  AFTER THE BURST: PLAY MODE - Picking: ray/triangle test against the skinned mesh, with a forgiving screen-space fallback for touch. - Grabbing pins the grabbed patch (radius ~0.4, only particles of that chunk) to a target on a camera-facing drag plane. Small chips move whole; big pieces stretch and swing like jelly. - Releasing throws the piece with the pointer's velocity. - Pieces collide with each other. A particle found inside another chunk's tet is pushed out through that chunk's nearest skin face, with friction. Use a chunk AABB broad phase and a spatial hash of skin tets. - Pieces stay on stage: side walls, plus an invisible front edge so nothing ends up under the controls or behind the camera. - The press still works: hold to squash the pieces again (no second fracture); Raise lifts the platen. - Wet "plop" sounds on landings; a small squelch on grab. - Cursor: open hand over pieces, closed hand while dragging. Dragging empty space orbits the camera.  UI (editorial, minimal) - Masthead top-left: "JELLY PRESS" in bold condensed caps, with "PRESS" filled in yellow/black hazard stripes. Subtitle: "Four gummies. One hydraulic press." - Top-right: Reset and Sound toggle. - Bottom deck:   - Caption line with escalating captions while pressing: "Contact." → "It's fine. It's jelly." → "Getting wider." → "That's a pancake now." → "It's making a noise." → "Please."   - Circular pressure dial (0–400 bar arc, red zone) around a red HOLD button, a Raise button, and a big numeric bar readout.   - Fruit chips with icons. - In play mode the caption slot shows "Grab a piece. Throw it." with small "Press again" and "Next jelly" buttons. - Controls: hold Space or ArrowDown to press, ArrowUp to raise, R to reset, 1–4 to pick a fruit. Wheel zooms; double-click resets the view. - Camera: low bench-level view; the press frame yaws per fruit so the posts never block the jelly. Framing adapts so the jelly sits between masthead and deck; works on phones (portrait) with a narrower stage.  SOUND (procedural Web Audio, no files) Hydraulic motor hum that rises with pressure, wet squelches, occasional creaks at high pressure, a valve clunk when the platen stops, a loud burst, soft plops on landing. Unlock on first interaction.  QUALITY BAR - Smooth 60 fps on a laptop. - Background warm-up of the other fruits' meshes and shaders so switching is instant. - Respect prefers-reduced-motion. - Accessible labels, a gauge with role=meter, focus-visible outlines. - No console errors. The page never goes blank.

TITLE: "SPARK"

An around 15-second 3D animated shot in the painterly style of the show Arcane (Fortiche).

Made in Blender, or any other tool if superior. Widescreen, no dialogue.

GOAL

The top priority is replicating Arcane's visual and animation style as perfectly as possible, in every way. Someone watching should believe it came from the same studio.

Take as much time and effort as needed. Replicate the technique and look perfectly. Spend as much time as needed, ensuring it's perfect in every way.

SCENE (loose — adapt freely)

A small mechanical creature (not humanoid), something like a brass-and-crystal moth, wakes up on a cluttered inventor's workbench at night. Its crystal core ignites with glowing energy, and it bursts into the air in a swirl of sparks. Change the details, framing, or action if something else shows off the style better. You can animate something else completely different if you want—anything you pick, whatever you can do best—that will also look identical to Arcane.

PROCESS

1. RESEARCH: Before building anything, study Arcane's style in depth. Find references and breakdowns of Fortiche's technique (interviews, making-of material, artist breakdowns). Write down every defining element: textures, shading, line work, color, lighting, frame rate, effects, camera, compositing.

2. STYLE GUIDE: Turn that into a written checklist and a small style frame (a single still image) before animating. Compare it side by side with reference stills and revise until it matches.

3. BUILD: Model, texture, light, and animate following the checklist.

4. REVIEW: Compare frames against Arcane references repeatedly. List every difference you can see and fix it. Repeat until no noticeable differences remain.

STYLE ELEMENTS TO MATCH (at minimum)

- Hand-painted textures with visible brushstrokes on every surface; nothing looks procedural or photographic.

- Stylized, painterly shading with designed light/shadow shapes, not realistic falloff.

- Animation on 2s for characters/objects, with strong poses, snappy timing, anticipation, and smear frames; camera moves smooth on 1s.

- Hand-drawn 2D effects (sparks, energy, smoke, glints, glow) layered over the 3D, animated on 2s with graphic shape language.

- Bold, moody color: warm light vs. saturated glowing accents, rich colored shadows, strong rim light, bloom.

- Painterly compositing: brush-like filtering, grain, subtle texture over the image.

- Cinematic camera: shallow depth of field, purposeful movement, weight on impacts.

SOUND

Detailed, cinematic sound design that matches the action and mood.

Make everything as good as possible: animation, models, textures, effects, lighting, and sound. Use any tools or programs needed. You may study references online and imitate techniques, and copy things, but do not directly use assets you did not create. Keep the whole shot in one consistent style so the 3D, painted textures, and 2D effects feel like a single hand-crafted image.

Do not use memory or previous chats.

You can use any other tools, programs, plugins, literally anything. Use anything at your disposal.

You can use different processes than outlined in this, or animate something else than described, but it should absolutely look as close as possible to the TV show Arcane. Make it perfect in every way and identical.

Create a beat-synced 3D sphere drop animation video with results comparable to a professional Blender production. Use multiple classic instrumental tracks, switching scenes with the music. As the 3D sphere falls, it should bounce off objects in each 3D scene, with objects lighting up in time with the beat. Include some humorous elements.
Use three.js, not Blender.

卡点的3D球体下坠的动画视频：对标专业的Blender做的效果，音乐是多首经典的纯音乐，场景按音乐切换，3D球体下坠时在3D场景的物件中弹跳，按音乐的节奏点亮物件。带一些幽默的元素。
使用three.js制作，不要使用Blender制作

Create "Jelly Press": a single-file interactive 3D toy in HTML (all JS, CSS and WGSL shaders inline, no external assets except Google Fonts). Render with WebGPU; if WebGPU or an adapter is missing, show a clean fallback message instead of a blank page.

CONCEPT
Four translucent gummy jellies shaped like fruit slices sit one at a time on the steel bed of a hydraulic press. The player holds a big red button to lower the press. The jelly squashes and spreads, the pressure gauge climbs, and somewhere past half its height it bursts into pieces. After the burst the game does NOT end: the player can grab the pieces, drag them, throw them around, and squash them again.

THE JELLIES (chips at the bottom, keys 1–4)
1. Watermelon wedge (half-disc slab): red flesh with dark teardrop seeds, a pale rind band, a green striped skin.
2. Orange slice (half-disc): orange pulp segments separated by thin white membranes, pale pith, orange peel.
3. Fig half: pink flesh full of small golden seeds, a cream layer, a dark purple skin.
4. Pineapple ring: golden fibrous flesh with radial streaks and a hole in the middle.
Each jelly should look like real gummy candy: subsurface scattering, soft translucency, glossy specular highlights, soft shadows on a warm studio floor (cream/beige, tone-mapped).

PHYSICS (CPU, fixed 60 Hz step)
- XPBD tetrahedral soft body with 8 substeps: co-rotational per-tet shape matching, per-tet volume constraints, hard edge strain limits (0.35×–1.8×), edge velocity damping, floor contact with Coulomb friction, rolling resistance, and gentle settling when nearly still.
- Render mesh skinned on the CPU through barycentric embedding in the tets; normals recomputed from triangles every frame.
- The press die is a kinematic round platen (radius ~1.05, rounded edge, a thickness, a ram above it). It acts as a ceiling with friction below, a shelf on top, and a side wall at its rim. The two press posts are solid.
- Pressure readout in bar comes from the platen's contact load, scaled per fruit.

THE BURST
- Break at a random squash between 52% and 66% of the jelly's height.
- Plan the fracture in the background shortly after each round starts, so the burst itself is instant.
- 5–7 big pieces from 3D Voronoi cells with slightly tilted walls. On 3–4 of them, a far corner is chipped off by two cutting planes and split further into 2–4 small chips, which leaves jagged, notched edges.
- Assign tets to cells by centroid. Duplicate particles per chunk. Merge tiny islands into their neighbours.
- The new body adopts the old positions and velocities.
- Clip the skin triangles against each cell's half-spaces, and fill every cut face with a clean flat cap showing the fruit's interior (flesh, seeds, membranes). No stretched triangles, no holes.
- Kick pieces outward and upward from the press. Small chips fly faster and higher and tumble with random spin.
- Show a big italic verdict for about 2.5 s, then fade it: "Splat." (melon), "Squeezed." (orange), "Well, that's jam." (fig), "Crushed." (pineapple). Include a stat line: "Gave up at N bar and N% of its height."

AFTER THE BURST: PLAY MODE
- Picking: ray/triangle test against the skinned mesh, with a forgiving screen-space fallback for touch.
- Grabbing pins the grabbed patch (radius ~0.4, only particles of that chunk) to a target on a camera-facing drag plane. Small chips move whole; big pieces stretch and swing like jelly.
- Releasing throws the piece with the pointer's velocity.
- Pieces collide with each other. A particle found inside another chunk's tet is pushed out through that chunk's nearest skin face, with friction. Use a chunk AABB broad phase and a spatial hash of skin tets.
- Pieces stay on stage: side walls, plus an invisible front edge so nothing ends up under the controls or behind the camera.
- The press still works: hold to squash the pieces again (no second fracture); Raise lifts the platen.
- Wet "plop" sounds on landings; a small squelch on grab.
- Cursor: open hand over pieces, closed hand while dragging. Dragging empty space orbits the camera.

UI (editorial, minimal)
- Masthead top-left: "JELLY PRESS" in bold condensed caps, with "PRESS" filled in yellow/black hazard stripes. Subtitle: "Four gummies. One hydraulic press."
- Top-right: Reset and Sound toggle.
- Bottom deck:
  - Caption line with escalating captions while pressing: "Contact." → "It's fine. It's jelly." → "Getting wider." → "That's a pancake now." → "It's making a noise." → "Please."
  - Circular pressure dial (0–400 bar arc, red zone) around a red HOLD button, a Raise button, and a big numeric bar readout.
  - Fruit chips with icons.
- In play mode the caption slot shows "Grab a piece. Throw it." with small "Press again" and "Next jelly" buttons.
- Controls: hold Space or ArrowDown to press, ArrowUp to raise, R to reset, 1–4 to pick a fruit. Wheel zooms; double-click resets the view.
- Camera: low bench-level view; the press frame yaws per fruit so the posts never block the jelly. Framing adapts so the jelly sits between masthead and deck; works on phones (portrait) with a narrower stage.

SOUND (procedural Web Audio, no files)
Hydraulic motor hum that rises with pressure, wet squelches, occasional creaks at high pressure, a valve clunk when the platen stops, a loud burst, soft plops on landing. Unlock on first interaction.

QUALITY BAR
- Smooth 60 fps on a laptop.
- Background warm-up of the other fruits' meshes and shaders so switching is instant.
- Respect prefers-reduced-motion.
- Accessible labels, a gauge with role=meter, focus-visible outlines.
- No console errors. The page never goes blank.

Act as a lead game designer. Build a complete, turn-based ASCII roguelike in a single self-contained HTML/JS/CSS file with zero external dependencies.
1. Visual: 1980s CRT monitor, phosphor green text (#00FF66) on black, soft scanline glow.
2. Procedural generation: 40x22 map, connected rooms and corridors.
3. Entities: @ hero, # wall, . floor, g goblin (5 HP), $ gold, > stairs down.
4. Mechanics: turn-based movement and combat, track HP and gold.
5. HUD: floor number, HP bar, combat log. Permadeath with restart.
Output only the working HTML code.

<create Game of thrones world>

Create a visually stunning 60-second video explaining how an earthquake happens through an interactive tectonic sandbox.

Make it feel like we are watching someone explore a beautiful real-time simulation, not a slideshow or traditional educational video.

Start with a clean 3D cross-section of Earth’s crust. Two tectonic plates slowly move against each other. Visualize the fault between them and show friction locking the plates while stress gradually builds.

As pressure increases, make the simulation more intense: rock layers deform, stress zones glow, subtle vibrations begin, and a live seismograph starts reacting.

Then trigger the earthquake. The fault suddenly slips and releases a massive burst of energy. Show seismic waves radiating outward through the ground, then transition upward to the surface where the landscape and a small city begin shaking.

Visualize P-waves and S-waves traveling differently through the Earth, followed by the strongest surface waves. Show buildings reacting differently depending on distance from the epicenter.

End by zooming back underground to reveal smaller aftershocks around the fault, then pull out to show the complete tectonic system.

Use cinematic motion graphics, satisfying physics simulations, dramatic scale transitions, premium 3D scientific visualization, minimal typography, dynamic labels, smooth UI overlays and seamless transitions.

The pacing should constantly evolve and reveal something new every few seconds so the full 60 seconds stays visually engaging.

Make it feel like an Apple-quality interactive science visualization turned into a cinematic motion-design video.

Create a mechanical Rube Goldberg machine in Blender. Use gears, ramps, balls, and moving platforms. Build it procedurally, inspect the scene after each major step, and fix obvious geometry issues.

I need you to launch five(Model) sub-agents and help me build a triple A quality game that is a clone of Mario Kart. What I want you to do is I want you to launch these sub-agents, build the game without asking me any questions at all, and use 3JS to build the game. And once you're done, report back to me.

Output a ThreeJS 3D cartoony/comics-like planet, with vibrant life on it: clouds, mountains, city buildings, cars and plane romaing it. It should be possible to orbit around and zoom and see the details. No harness is to be used

Create a stunning 30-second motion graphics video for https://t.co/71rvEGmB6D that feels like an elite motion designer’s showreel. Study the website first and use the real brand, product UI, colours, and messaging. Use bold typography, 3D motion, animated UI, fast transitions, and a strong logo reveal.

Create an extraordinary interactive SVG experience titled “EARTH ATLAS: THE LIVING PLANET.”  Build a complete, deeply explorable Earth science observatory in ONE self-contained HTML file. I must be able to paste your entire output into a file, open it directly in Chrome, and use it without a server or network connection.  Use whatever libraries help, but bundle all required runtime code and curated data inside that one HTML file. The globe, regional maps, terrain, seafloor, cutaways, instruments, charts, annotations, and close-up illustrations must be rendered primarily in SVG.  Return ONLY the complete working HTML in ONE code block.  THE AMBITION  Create the visual quality of a flagship NASA Earth-science exhibition combined with the depth of an interactive atlas.  The viewer begins in orbit before a magnificent Earth. They can rotate the globe, move through daylight and darkness, choose a real place, zoom into its geography, draw a transect, descend through its surface or ocean, and understand how that location fits into the whole planet.  The experience must reward ten minutes of exploration. It should contain striking reveals, precise interactions, and connected views—not merely a rotating globe with informational popups.  The first frame must be beautiful enough to serve as a standalone scientific illustration.  A TRUE MULTISCALE EXPLORATION SYSTEM  Build an authored hierarchy of scales:  1. ORBITAL EARTH — the full globe, atmosphere, day–night boundary, large geographic forms, and global overlays. 2. CONTINENTAL — recognizable coastlines, relief, major rivers, ocean basins, and selected scientific features. 3. REGIONAL — detailed terrain, bathymetry, contours, labels, and measurement tools. 4. LOCAL SCENE — a richly illustrated landscape or ocean environment specific to the selected place. 5. SECTIONAL VIEW — a geographically linked cross-section through land, ice, ocean, or crust. 6. DETAIL VIEW — close examination of a feature such as a glacier layer, river channel, fault, sediment bed, or seafloor formation.  Each zoom level must reveal newly authored SVG detail appropriate to its scale. Do not simply magnify the same low-resolution geometry until it becomes empty.  Preserve geographic identity across transitions. A selected point on the globe must correspond to the regional map, the local scene, its transect, and its section.  Implement zoom toward the pointer, drag-to-rotate the globe, drag-to-pan regional views, pinch zoom, keyboard navigation, Back, Home, and Reset View. Keep camera movement bounded and smooth.  Show a discreet scale bar and location breadcrumb. Manual interaction must interrupt any automated camera journey gracefully.  VISUAL DIRECTION  Make Earth the dominant visual element.  Use deep ocean blue, shallow-water turquoise, varied vegetation greens, desert ochre, volcanic charcoal, mineral grays, glacial white, and fine warm-white annotation lines.  Build convincing planetary volume through projected coastline geometry, changing illumination, atmospheric scattering effects, cloud layers, and a carefully controlled night-light treatment where an appropriate dated source is embedded.  Use precise cartographic linework, elegant contours, subtle material patterns, readable depth shading, and thin leader lines. Make information appear when useful at the current zoom level rather than covering the globe with labels.  The interface should feel like a refined scientific instrument. Avoid generic dashboard cards, excessive neon, and large floating text panels.  GLOBAL EARTH  Provide a rotatable SVG globe with consistent geographic projection and coordinate handling.  As the globe turns, correctly project landforms, ocean features, labels, selected markers, routes, and the terminator. Hide geometry on the far side rather than drawing it through the planet.  Include switchable overlays for:  • Physical relief. • Ocean depth and seafloor structure. • Major rivers and drainage basins. • Plate boundaries. • Selected historical earthquake events from an embedded dated sample. • Latitude–longitude grid. • Dated Earth-observation samples, only where embedded data supports them. • Clean cinematic view.  Every overlay must use the same coordinate system. Include its source, date, resolution, units, and legend.  Do not call any layer “live” or “current” when the file uses an embedded snapshot.  SIX FULLY AUTHORED EXPEDITIONS  Build complete, distinct experiences for:  HIMALAYA Layered mountain ranges, valleys, glaciers, a sourced regional elevation profile, and a visually clear section through the terrain.  AMAZON BASIN Recognizable river structure, floodplain and tributaries, an illustrated forest cross-section, and an educational water-path demonstration.  SAHARA Dunes, rocky terrain, dry channels, distinct surface materials, and a terrain transect. Distinguish illustrated dune detail from sourced large-scale elevation.  EAST AFRICAN RIFT Regional lakes and terrain, a fault-oriented cross-section, and a clearly labeled conceptual tectonic demonstration.  MARIANA TRENCH REGION Ocean surface, shelf and deep-ocean geometry where appropriate to the chosen transect, a draggable depth profile, a descent through the water column, and a detailed illustrated seafloor scene.  ANTARCTICA Ice surface, a sourced ice/terrain context where available, a sectional interpretation, and a seasonal sunlight demonstration.  Give every expedition:  • A strong authored opening composition. • Correct placement on the globe. • A regional map. • At least one local scene. • At least one linked cross-section or profile. • A unique working interaction. • A smooth journey back to orbit.  Do not reuse a generic mountain, forest, desert, or ocean drawing as a substitute for geographic identity.  FIVE CONNECTED VIEW MODES  SURFACE Explore topography, terrain features, rivers, ice, and coastlines with zoom-dependent detail.  ABOVE Explore sunlight, seasonal geometry, atmosphere, and any embedded dated observation. If clouds or weather are illustrative, label them as an illustrative scenario.  BELOW Open a section through terrain, ice, crust, or ocean along the selected geographic transect.  PLANET Pull back to a global cutaway showing how the selected point relates to Earth’s layers. Show the selected location’s corresponding position on the globe and cutaway.  EVIDENCE Reveal the data sample, source, date, resolution, uncertainty or known limitation, and calculation behind the selected feature or measurement.  Changing modes must preserve the selected location, zoom context where sensible, and timeline state.  LINKED EXPLORATION TOOLS  MEASURE Select two points on the globe or a supported regional map. Draw the great-circle route, display coordinates and distance, and keep the route correctly projected during globe rotation.  TRANSECT Draw or adjust a line across a supported region. Sample its embedded elevation or bathymetry data to produce a profile. Moving a cursor on the map must move its counterpart on the profile and sectional view.  COMPARE Place two selected locations side by side. Use matching units and explicit scale controls. Compare their elevation or depth, latitude, regional context, and supported environmental data.  TIME AND SUNLIGHT Scrub time of day and day of year. Update the global illumination and seasonal sun geometry with a documented model. Keep illustrative weather independent of this clock.  GUIDED EXPEDITION Offer a short cinematic journey through the six environments. Each stop must reveal an interaction, not merely display a caption. The tour must be skippable.  ANNOTATION DENSITY Switch between cinematic, guided, and technical label levels without altering the underlying geography.  OCEAN EXPLORATION  Treat the seafloor as a complete landscape.  Reveal shelves, slopes, abyssal regions, ridges, and trenches where supported by the embedded relief sample. Use an explicit vertical scale and label any exaggeration.  In the Mariana expedition, let the viewer descend through the water column. Update depth, light, color, pressure estimate under a stated simplified model, and the location cursor together.  The deep-sea close-up may contain beautiful illustrated life and geology, but its exact organisms and microterrain must be identified as interpretation. Sourced bathymetry and illustrated local scenery must remain distinguishable.  EARTH’S INTERIOR  Provide a 0–100% cutaway slider that smoothly reveals the crust, mantle, outer core, and inner core. The globe should remain visually coherent at intermediate slider positions.  Show the selected geographic location on the outer surface and an aligned radial guide into the interior.  Include an educational seismic-wave demonstration with a clearly stated simplified model. Its wave paths, moving markers, and timing readouts must derive from the same demonstration state.  Label exaggerated layer thickness, material colors, or compressed time explicitly.  THREE INTERACTIVE SCIENCE EXPERIMENTS  1. SUNLIGHT LAB Choose two latitudes and compare the modeled daily solar path and length of daylight across the year. Link the diagrams to the illuminated globe.  2. RELIEF AND SEA-LEVEL LAB At a supported coastal region, adjust a hypothetical water level and compare it with the embedded elevation profile. Label this a static topographic demonstration; do not present it as a coastal flood forecast.  3. SEISMIC PATH LAB Choose a source and observation points for a simplified educational wave demonstration. Show paths and relative arrival timing according to the stated model. Keep historical earthquake markers separate from the hypothetical experiment.  Each experiment must have Reset, reproducible inputs, consistent units, and a clear link back to the globe or selected place.  DATA AND SCIENTIFIC HONESTY  Use authoritative, cited sources for numerical claims and embedded samples. Appropriate sources include NASA Earthdata for selected dated observations, NOAA ETOPO for land and ocean relief, and USGS for selected historical earthquake records.  Use curated, downsampled data that fits inside one HTML file. Show the actual embedded resolution. Never claim that a regional illustration has the precision of the source dataset when its detail was authored or simplified.  Keep three categories visible:  OBSERVED DATA — a sourced measurement or published dataset sample. DERIVED VALUE — calculated from named inputs and an inspectable method. ILLUSTRATION OR EXPERIMENT — authored scenery or a hypothetical model.  Do not invent precise elevations, depths, earthquake positions, live cloud patterns, or measured environmental values.  Include a Sources and Methods panel with links, dataset versions and dates, coordinate reference, units, downsampling method, equations, uncertainty or limitations, and attribution.  Use a consistent latitude–longitude convention across every feature. Keep geographic coordinates, physical measurements, and exaggerated display geometry separate.  SMALL DETAILS AND VISUAL REVEALS  Include carefully choreographed discoveries:  • Orbit-to-region transitions that keep the chosen point in view. • Contours and labels that emerge only when the viewer reaches their useful scale. • A river that remains recognizable from global map to local basin. • An elevation cursor moving in synchrony across map, profile, and landscape. • A coastline that transforms visually when the bathymetry layer activates. • Ice layers that reveal themselves progressively in the Antarctic section. • A deep-ocean descent in which the seafloor appears gradually rather than as a sudden scene swap. • A globe-to-interior cutaway that preserves the selected geographic location. • A changing terminator that casts the chosen expedition into daylight or night. • An evidence reveal connecting a beautiful visual element to the data or model behind it.  Prioritize details that deepen exploration over decorative particles.  PERFORMANCE AND VERIFICATION  Use a bounded SVG scene graph, reusable symbols, clipping, masks, and zoom-dependent rendering. Avoid keeping all six high-detail scenes active when only one is visible.  Support desktop, touch devices, keyboard navigation, visible focus, and reduced-motion preferences.  Verify that:  • Every featured location is geographically placed correctly. • Hidden-side globe features are actually hidden. • Routes remain attached to their coordinates through rotation. • Measurement distances use the selected coordinates. • Transect values come from the embedded samples. • Map, profile, and section cursors remain synchronized. • Reset returns experiments to their initial states. • Time controls alter the intended model without silently modifying unrelated data. • Source and date labels match the embedded layers.  Show actual results for any automated checks. Do not hardcode a row of “passed” labels.  Make the complete path from globe to local scene to section work for all six expeditions. If implementation scope forces a tradeoff, complete that connected journey and the essential science tools before adding optional visual effects.  FINAL DELIVERY  The finished result should feel like an explorable planet inside a single file: stunning from orbit, rewarding at close range, and clear about what is observed, calculated, or illustrated.  Return ONLY ONE code block containing the ENTIRE working HTML document, beginning with <!DOCTYPE html>.  I should be able to paste it into one .html file, open it in Chrome, rotate and zoom Earth, enter any of the six expeditions, draw a linked transect, explore the ocean and interior, compare locations, run the three experiments, inspect the sources, and return smoothly to orbit.

Build an animated, interactive discrete-event simulation of a serial production line as a single self-contained HTML file (vanilla JS + Canvas, no external libraries except Chart.js from cdnjs). Purpose: teaching [Manufacturing Systems] students how variability, buffers, and breakdowns affect line performance.

LINE CONFIGURATION (user-adjustable)
- Number of stations: [2–6], default [3], each with a name (e.g., Machining, Assembly, Inspection)
- Per station: mean cycle time, distribution (Deterministic, Uniform, Normal, Triangular, Exponential, Lognormal), CV
- Breakdowns per station: MTBF and MTTR (exponential), on/off toggle
- Quality per station: defect rate (%), with scrap or rework-loop option
- Buffers between stations: capacity 0–10 (0 = blocking after service)
- Arrivals: unlimited raw material OR Poisson arrivals with rate λ
- Customer demand to compute takt time

ANIMATION
- Parts move along conveyors and change color by stage (raw, WIP, finished, scrap)
- Machine borders show state colors: Working (green), Blocked (amber), Starved (red), Down (gray), plus a rotating gear icon and a progress bar
- Buffer slots show occupancy and highlight when full
- Controls: Play/Pause, Step, Reset, speed 1x–50x, warm-up period

KEY PERFORMANCE MEASURES (live dashboard)
1. Throughput (parts/hr) vs theoretical bottleneck rate
2. Average WIP and WIP over time
3. Flow time / manufacturing lead time (mean and 95th percentile)
4. Little's Law check: WIP ≈ Throughput × Flow time
5. Per-station utilization with a time breakdown: working / blocked / starved / down (stacked bar)
6. OEE per station = Availability × Performance × Quality
7. Bottleneck detection (active-period method) and highlight the bottleneck
8. Takt time vs station cycle times (line balance chart)
9. Line balance efficiency = Σ cycle times / (N × max cycle time)
10. First-pass yield, rolled throughput yield, scrap count
11. Average buffer occupancy per buffer
12. Observed vs set cycle-time mean and CV per station

ANALYSIS FEATURES
- Replication mode: run N replications of length T after warm-up, report mean ± 95% confidence interval for throughput, WIP, and flow time
- Experiment mode: sweep buffer size (or one station's CV) and plot throughput vs parameter
- Export results to CSV
- Preset scenarios: Balanced line, Clear bottleneck, High variability, Unreliable machine

DESIGN
- Clean, responsive layout that works on mobile; light/dark mode
- Brief tooltip explaining each KPI and its formula
- Bilingual labels (English/Arabic) toggle

Create a walking building.

歩く建築作って

Using the Magnific MCP, build a playable pixel-art beat 'em up set in ancient Rome, as one HTML file that works on mobile. Generate a key art first and use it as style reference for every asset: stage, hero, enemies, a war-elephant boss, items. Animate the characters with image-to-video on green screen, pick the frames that loop, key out the green and keep every animation on the same scale and palette. Add touch controls, combos, shield, dodge, a throwable pilum, pickups, a 1-minute demo mode and a chiptune soundtrack. Tell me the credit cost before each generation.

Build immediately a complete interactive WebGPU site called “Strawberry Cake.” 
Use genuine WebGPU + WGSL, procedural geometry, mouse/touch, and real-time volumetric soft-body physics. No Three.js/Babylon.js, Canvas2D, external assets, video/GIF, or CSS-only deformation.
Visual: premium Korean/Japanese strawberry shortcake—low, wide, rounded, pillow-soft, with pink sponge, cream layers, pale frosting, piped cream, and strawberries. Warm ivory background, soft studio light, edible moist materials.
Layout: top-left “SOFT STUDIES / NO.001” + italic “Strawberry Cake.” Top-right “WEBGPU · LIVE”. Large centered cake. Right controls: Hand/Knife, presets, Firmness, Damping, Drop, Reset, Pause, ¼ speed, Show mesh. Bottom-left: Mass, Volume, Kinetic, Pieces.
Physics: stable XPBD/co-rotational soft body with tetrahedral sim mesh, smooth render mesh, volume preservation, damping, gravity, floor friction, and piece collisions. Cake feels soft/heavy and wobbles after release.
Hand: hold=Press, inward drag=Squeeze, outward drag=Grab, fast release=Throw. Use weighted grabs, smoothed 3D target, pointer velocity and momentum. Cake remains free; cake input overrides orbit.
Knife: procedural 3D knife. Draw a cut line; animate contact→compression→penetration→breakthrough→lift. Deform before splitting; keep seam thin.
Repeated cutting is mandatory: use a dynamic pieces list; every piece stays cuttable. Convert strokes to vertical cut planes, split crossed pieces, create new soft-body/render meshes, transfer deformation/velocity, preserve toppings, keep halves aligned, update Pieces. Support 14+ pieces and multi-piece cuts. No fixed whole/left/right states.
Drop releases all pieces. Reset restores one intact cake and Pieces=1. Use fixed timestep/substeps, DPR≤2, clamp instability, avoid NaNs/GPU errors. Expose window.__cake.
Run locally, test interactions and repeated cutting, and iterate until polished.

I want to build a lego set of the original ford motel T.

I want it to be feature complete, include motion/interactivity where possible, and be of the quality a lego master builder would be proud of.

I want the final output to include a rendering of the lego set, all of the pieces i need to order from lego, and the instruction manual to build it.

Build me a hyper-realistic multiplayer FPS. Set it in a snowy city alleyway with brick buildings. Give me an assault rifle with crispy recoil and flying bullet casings. Add enemies to shoot, vaulting mechanics, and bloody screen damage effects. Make the gunplay hit-reg immaculate and just vibe it out into a full game.

Give me a 55 second 3D scene. Camera flies into a data center, opens a rack, disassembles a GPU, zooms into the chip, passes through the transistors, lands on a single silicon atom. Scale rail on the right in meters. Labels in the client's language. Runs in a browser, one file, no dependencies.

make a dynamic 30-second motion graphics video that shows what an incredible motion designer you are for https://t.co/fCRvqmOamH.

Could you also create an anime-style 3D CG scene, like a robot anime, where three vehicles transform and combine to become a robot?

もしかしてロボットアニメみたいな3機の乗り物が変形合体してロボットになるシーンもアニメ風3D CGで作れますか。

Ask me for: the task the AI agent completes for the user (default: plan + book a 3-day Kyoto trip), the palette (default: warm-gray canvas #E6E3DE, pure black/white UI, the orb is the only colored thing — pearl iridescent blue → violet → peach), and a royalty-free song around 120 BPM (search Mixkit yourself, measure BPM with numpy, show me 3 candidates with preview links before downloading the full track). Use AskUserQuestion, max 4 questions per round, recommended option first. Dribbble-level UI concept motion for an AI chat agent tool. One white shape, never cut: every state is the same element morphing its size, radius and color while its content swaps with a short blur. The star is a GLSL "AI orb" (fluid, breathing, iridescent blob — like the ChatGPT / Siri voice orb) that lives OUTSIDE the shape as a continuity element and changes role through the film: idle hero → reacts to voice while listening → shrinks into the input bar as an avatar → swirls faster while thinking → avatar on the answer → blooms when the task completes → back to idle. A cursor drives every change with real clicks and drags. One clean UI font (Geist). Springs everywhere, a tiny overshoot at most. The camera zooms so each state fills ~60–75% of the frame. The orb pulses subtly on every beat. The last frame is the first frame, so it loops. Banned: bouncy easing, particle bursts, glows or gradients on UI chrome (the orb is content, not chrome), mismatched icon strokes, dead time, anything that looks like a template. 120 BPM, 8 bars = 32 beats = 16s, something happens on every beat (beat n at (n-1)*0.5s): Bar 1: orb idle, cursor approaches | click orb → shape stretches into a "LISTENING" pill, orb moves to its left and wobbles with a syllable envelope | live transcript types "Plan 3 days in Kyoto" | "…under 1,500, the value rolls live | drag past the 1,248 "cheapest plan" Bar 6: slider → black swipe-to-confirm "Book trip · $1,248" | grab the knob, drag right | past the end → rubber band | release → snaps to the end, arrow becomes a spinner Bar 7: → toast "Trip booked" | chips "Flights ✓" "Ryokan ✓" pop in | orb blooms (color + scale) | toast collapses toward the orb Bar 8: shape fades, orb grows back to hero size | settles | breathes on the beat, cursor drifts out | back to idle (loop) 1. One self-contained HTML file, square 1440x1440, fonts and audio-free assets inlined as data URIs. Every style is computed from time inside a pure `seek(t)`: no CSS transitions, no timers, no state carried between frames, never create tracks inside seek. 2. Springs are closed-form step responses. A value that changes target many times is the sum of one spring per change; to loop, take the LAST target as the start value and also add the spring tails of the previous two cycles (t + L, t + 2L) so position AND velocity match at the seam. Damping ratio ≥ 0.72. 3. Content layers have their own enter/exit windows: exit lands exactly on the beat, enter starts ~80ms later (opacity + ~12px screen blur + 0.965→1 scale), or text overlaps. 4. Drags are direct manipulation: while held, the value = start + (cursorX − cursorX at press); past a limit apply rubber(over, R) = R·(1 − e^(−over/R)); on release a free spring runs from the release position AND velocity to the snap target. Cursor waypoints are [departure time, x, y] on a spring; the final waypoint equals the first so the cursor's position and speed are continuous across the loop. 5. The orb: WebGL fragment shader drawn on an OFFSCREEN 640×640 canvas, then synchronously drawImage'd into a visible 2D canvas inside seek (screenshotting the WebGL canvas directly is unreliable headless). Time must be periodic: uniforms (cos, sin)(2π·k·t/L) with integer k (e.g. k=2 and k=5), never raw t. Silhouette radius = 0.74 + amp·noise(direction·1.4 + T) sampled on the unit direction vector (no center seam); surface = low-frequency domain-warped 3D noise on the sphere normal → large smooth color bands; 16% pearl white, a bright inner core, a small sharp specular (pow 70), a pale lilac fresnel rim; darker lower half for volume; soft elliptical shadow under the idle orb. Uniforms: amp (idle 0.08, + syllable envelope while listening, + a bit while thinking), think (stronger/faster warp), bloom (completion color burst). Keep ≤5 noise calls per pixel. 6. Audio: analyze the song with numpy (spectral-flux onset, phase-locked BPM, downbeat by kick + chroma change, per-bar RMS). Start on the downbeat of a full-energy 8-bar phrase so the audio loop lands on a phrase boundary. Large analysis windows estimate beats ~15–25ms early: re-measure the cut with a 256-sample window and shift the start until the median residual is < 2ms. Crossfade the last 60ms with the 60ms before the start. Synthesize UI sounds with numpy (click, send swoosh, grab/drop, tick, success, chime) and place each by its MEASURED PEAK on the event time; multi-note sounds keep the first note loudest. 7. Render with Playwright Chromium: 4 subframes per frame spread over half a frame (180° shutter, centered on the frame time), 16 parallel workers. After each seek, await two requestAnimationFrames before the screenshot; do NOT use screenshot(animations='disabled'). Because the scene uses WebGL, launch with --use-gl=angle --use-angle=swiftshader --enable-unsafe-swiftshader --disable-gpu-watchdog (the default SwiftShader-Vulkan path loses the WebGL context most of the time). Convert any RGBA screenshot to RGB. 8. Encode with ffmpeg: -reinit_filter 0, tmix=frames=4 then select every 4th frame, 60fps, libx264 -crf 10 -x264-params aq-mode=3, AAC 256k, -t exactly the duration. Verify which subframes tmix actually averages on your ffmpeg version before choosing the select offset. Also make a crf 20 share version and a 3× looped copy. 9. Before the full render: render one frame on each beat and one 0.3s after it, tile them into contact sheets, and fix anything off the grid, cramped, clipped, unreadable, or with the orb missing. Then spot-check transition frames at full resolution. 10. Verify and report: t=0 and t=L screenshots are pixel-identical; frame count = 960 and video/audio are exactly 16.000s; seam frame-diff is the same order as its neighbors; no RGBA subframes; beat offset of the final audio < 10ms; sample the idle frames to confirm the orb is present in every one. - smoothstep(e0, e1, x) with e0 > e1 is undefined in GLSL — SwiftShader returns 0 and the orb vanishes. Always write 1.0 - smoothstep(lo, hi, x). - A heavy shader on software GL triggers the GPU watchdog (CONTEXT_LOST_WEBGL): keep the canvas at 640² and the noise count low; listen for console context-loss messages in every test. - Never put will-change on anything the camera scales, or text renders blurry. - Followers (the orb, avatars) use a slightly slower spring than the shape so they never get clipped by its edge. - Layer order: anything with a background plate goes BEFORE the text/icons that sit on it. - Don't override a centering translate() with an animated transform — wrap it in a layer instead. - ffmpeg -shortest can drop the last frame; set -t explicitly. - Make the last frame identical to the first, cursor position and speed included, or the loop stutters. Ask me for the inputs, prototype the orb shader alone first (render t=0, t=4, t=8 and prove t=0 == t=16 pixel-for-pixel, with no context loss across 8 fresh page loads), then show me the state list on the 8-bar beat grid as a table and wait for my OK before writing the full scene.

I’m curious to see how strong you are as a motion designer. To prove it, create a polished 15-second animated graphic video that showcases your motion design skills. This piece will be presented to 10,000 people on X, so make it visually compelling, professional, and memorable. Take full creative freedom and make the strongest result you can

Build a Minecraft-style voxel game in a single HTML file that runs in the browser.
- First-person controls: WASD, mouse look, jump
- Procedural terrain with hills, water and trees
- Place and break blocks with the mouse, 5 block types
- Advanced shaders: moving sun, soft shadows, ambient occlusion, fog, water reflections
- Smooth 60 fps on a laptop
Use Three.js from a CDN. Test it, fix every bug, then keep improving the visuals until it looks as realistic as possible.

Create a complete, polished Spotify-themed motion graphics video from the following brief. This prompt is self-contained: no reference video or supplied assets are required.

<deliverables> An 18-second MP4 at 1920×1080 and true 60 fps, with music and subtle sound effects Complete editable source Build, render, inspect, and refine the animation before delivering it Do not stop at a storyboard, still images, or an implementation plan </deliverables>
<art_direction>
A premium music-product film presented inside a large, nearly black rounded rectangular stage
The stage occupies roughly 80% of the canvas width and 75% of its height, centered horizontally and slightly below the vertical center
Outside the stage, use a soft atmospheric green background: luminous emerald near the upper-left corner, deeper forest green toward the sides, fading to near-black at the bottom
Inside the stage:
Near-black background
White primary typography
Muted gray secondary text
Spotify green # 1ED760 for emphasis and controls
One clean sans-serif font, such as Geist or Inter
Consistent icon strokes and restrained shadows
Keep compositions compact and centered, with substantial negative space
Text should feel like music-product advertising, not oversized presentation headings
Do not add an outer caption, creator watermark, decorative footer, or progress counter
</art_direction>

<artwork> Create original square album artwork for a fictional track called “Glass Tides” by “AURA”
The hero cover is an abstract macro image of flowing pearlescent liquid silk and molten glass
Use pastel lavender, powder cyan, blush pink, champagne gold, and subtle mint
Include broad three-dimensional folds, realistic reflective highlights, delicate striations, and a prominent sweeping S-shaped fold
Fill the square edge to edge
No text, logo, or border on this hero artwork
Use this same artwork consistently throughout the film

Create additional designed playlist sleeves:
“Late Nights” in mustard yellow
“Good Energy” in pink
“Deep Focus” in blue
“Daily Mix 1”, “Daily Mix 2”, and “Daily Mix 3” in complementary colors

These should look like finished graphic-design covers, with bold typography and simple geometric motifs
</artwork>

<timeline> 0.0–0.6 seconds: A green Spotify icon grows smoothly into the center of the dark stage Give it a confident arrival with minimal overshoot
0.6–1.5 seconds:
The icon transitions into a small Spotify identity above two centered lines:
“Discover”
“new music”
The first line is white; the second is green
Reveal them with short masked vertical movements
1.5–2.4 seconds:
A stylized Spotify desktop interface rises into view
Show a slim sidebar, “Made for you”, three album cards, and small supporting rows
Use a subtle perspective tilt during the entrance, settling toward a frontal view
2.4–3.2 seconds:
Move closer to the three featured covers
Show “New for you” above them
Keep the artwork crisp, with short album titles beneath
A narrow rounded player bar now anchors the bottom of the stage
3.2–4.0 seconds:
The featured cards withdraw into a compact “Fresh finds” list
Reveal four song rows with a slight stagger
Each row contains a thumbnail, short track title, artist, and a small menu icon
4.0–4.9 seconds:
Transition into a horizontal strip of colorful playlist covers
Heading: “Every mood”
Supporting line: “Find what moves you”
The strip slides smoothly sideways, with edge cards partially cropped by the stage
4.9–5.8 seconds:
A compact search field appears
Type “Glass Tides”
Reveal one selected result beneath it, with the iridescent artwork, “AURA”, and a green play icon

5.8–6.6 seconds:
The selected artwork expands into a large centered cover
Reveal “Glass Tides” underneath, followed by “AURA” in green
Keep the player bar visible
6.6–7.5 seconds:
The same cover slides left
Track information appears on the right:
“Glass Tides”
“AURA”
“A new frequency”
Add a rounded green Play button
Coordinate the cover movement and text reveals
7.5–8.3 seconds:
A brief brand-color transformation fills the inner stage with green
The artwork contracts into a smaller centered tile
Place a small dark Spotify identity underneath
The player briefly recedes
8.3–9.8 seconds:
Return to the dark stage
The same artwork expands dramatically toward the camera, becoming oversized and cropped by the rounded stage
Restore the player at the bottom
Use controlled motion blur during the fastest part of the zoom

9.8–11.2 seconds:
The artwork clears into a centered typographic statement:
“Find your”
“rhythm”
Use white for the first line and green for the second
The player remains visible and stable beneath it
11.2–12.5 seconds:
Show “Made for your every day”
Three playlist sleeves enter with gentle perspective and small opposing tilts
They spread into six smaller sleeves across the stage
Supporting line: “Your sound, always evolving”

12.5–14.0 seconds:
The playlist layout transitions into four floating album covers
Include the iridescent hero artwork
Use restrained rotation, perspective, depth ordering, and overlap
Keep their movement coordinated rather than randomly floating
14.0–15.2 seconds:
The covers converge toward the center and fold around a green Spotify icon
The artwork withdraws as the icon becomes the focal point
Fade the player away
The Spotify icon must emerge from the same central position as the converging covers
15.2–18.0 seconds:
The icon moves slightly left and settles beside a large white “Spotify” wordmark
Keep the icon and letters separated throughout the movement
Reveal:
“Discover new music”
“every day”
The first line is white; the second is green
Hold the finished composition cleanly through the end
</timeline>
<persistent_player>
For most of the middle sequence, keep one narrow rounded player bar near the bottom of the inner stage
Include:
Small hero-art thumbnail
“Glass Tides” and “AURA”
Play and skip controls
A thin progress track
A small volume icon
Subdued dark translucent styling with a fine border
This recurring player connects the changing shots
It must remain visually secondary to the artwork
</persistent_player>

<motion_quality>
Match the energy of a tightly edited premium product film
Most visual ideas last approximately one second, but transitions remain smooth
Use continuous acceleration and deceleration
Favor critically damped springs or carefully tuned smooth easing
No repeated bouncing or large elastic overshoots
Preserve the hero artwork’s identity and position relationships through search, album detail, brand tile, and zoom
Use match-position transitions, coordinated scaling, masked reveals, and perspective
Outgoing titles must disappear before incoming titles occupy the same space
Avoid overlapping text, sudden camera resets, long blank intervals, and arbitrary full-frame crossfades
Clip every oversized cover and camera move cleanly to the rounded stage
No particles, shockwave rings, lens flares, camera shake, or unrelated stock footage
</motion_quality>

<audio> Create or select commercially usable electronic music around 120 BPM Use a clean pulse, warm bass, restrained melodic elements, and subtle transition accents Align important entrances, selections, zooms, and the logo reveal with musical events Keep effects quieter than the music No voiceover Do not use copyrighted commercial tracks without permission </audio>
<implementation_and_validation>
Build a deterministic animation driven by absolute time through an async window. seek(t) function
Every transform, opacity, mask, and UI state must be reproducible when seeking frames in any order
Do not depend on live timers, accumulated physics, or CSS transition state during export

Render true 60 fps with spatial antialiasing
Use 3–5 temporal subframe samples per output frame for restrained motion blur
Keep stationary text and artwork sharp
Inspect contact sheets and moving playback
Check the fast heading changes, artwork handoffs, zoom, cover convergence, and final wordmark spacing
Verify evenly spaced frame timestamps and the full 18-second duration
Fix visual defects before delivering the final MP4 and editable source
</implementation_and_validation>

make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out.

Build a browser sandbox game in the spirit of Minecraft that feels as close to the original as possible. All in-game text in English. Controls: keyboard and mouse (desktop).  TECH - A single HTML file, plain WebGL2, no third-party libraries. - All 16×16 textures generated in code as pixel art (stone, dirt, grass, planks, leaves, ores, glass, water, lava, etc.). - Sounds synthesized with WebAudio: digging, footsteps, placing blocks, damage, mobs, explosions, calm background music.  WORLD - Infinite world made of 16×16×128 chunks, with a seed. - Biomes: plains, forest, birch forest, taiga, snowy tundra, desert, mountains, oceans, beaches. - Caves (winding tunnels and large caverns), lava at the lower levels, ores by depth: coal, iron, gold, diamonds. - Three tree types, tall grass, flowers, cacti, sugar cane, pumpkins. - Minecraft-style lighting: sky light and block light (torches, glowstone, lava) that spreads cell by cell, with smooth lighting and ambient occlusion. - Day/night cycle: sun, moon, stars, sunsets, 3D clouds, distance fog, rain. - Water and lava flow by levels; two sources make infinite water; water + lava make obsidian or cobblestone. Sand and gravel fall.  PLAYER - First person, collisions, jumping, sprinting, sneaking (doesn't fall off edges), swimming, ladders, fall damage. - Block breaking with crack stages and particles; break time depends on the tool. - Visible hand and held item with a swing animation. Third-person view on F5.  SURVIVAL - Health, hunger, saturation, air underwater. - Tools in 5 materials with durability, armor in 4 materials. - Inventory with 2×2 crafting, 3×3 crafting table, furnace with fuel, chests, bed (skip the night and set spawn). - Item drops, death and respawn. - Mobs: pigs, cows, sheep, chickens (breeding, shearing); at night zombies, skeletons with bows, spiders. Zombies and skeletons burn in sunlight. - Farming: hoe, seeds, wheat growth, bread. Doors, fences, fence gates, TNT.  CREATIVE - Flying on double-tap Space, instant block breaking, a catalog of all blocks with tabs and search.  UI - Title screen with a world panorama, world list (create / delete / play), options (FOV, render distance, sensitivity, sound, brightness, GUI scale). - Pause menu, death screen, HUD (hotbar, hearts, hunger, armor, air bubbles), F3 debug screen. - Chat with commands: /gamemode, /time, /give, /tp, /summon, /weather. - Worlds saved in localStorage.  RESTRICTIONS - Do not use Minecraft's name, logo, textures or characters (Steve, Creeper, etc.): give the game its own name and design your own mobs.  TESTING - Run the game in a headless browser, check every system and fix bugs before delivering.

Implement code to be able to navigate in a pagoda in 3D.

Please use Three.js to build a "freely walkable 3D anime-style cherry blossom town" as a single HTML file, then publish it as a shareable web page.

[Technical constraints]
- Use only three.js r128 (UMD build) from cdnjs. Do not load any external models or images — generate all models, textures, and shop signs procedurally with code and Canvas.
- Use original content for all shop names, signs, and characters. Do not imitate any real brand or existing work.
- Use MeshStandardMaterial or MeshPhongMaterial. Avoid high metalness and environment reflection maps (on some computers they make objects render with no color).
- Merge static objects by material into a small number of meshes so it runs on ordinary computers; provide a High / Medium / Low quality switch.

[Scene: "桜ヶ丘 (Sakuragaoka)", a small Japanese town on a spring afternoon]
1. Shopping street: a north–south main street with 20+ shops on both sides (ramen shop, café, bicycle shop, bookstore, florist, Japanese sweets shop, drugstore, convenience store, etc.). Each shop has: a multi-line sign (shop name + English name + phone number), a striped awning with a scalloped valance, a recessed storefront with an interior that has visible depth, and a sidewalk display (fruit crates, magazine rack, food-sample case, a spinning barber pole). Upper floors have windows, AC units, balconies with hanging laundry, and rooftop TV antennas.
2. Street details: utility poles with lots of overhead wires, decorative street lamps with shopping-street banners, festival lantern strings, square sidewalk tiles with yellow tactile paving, drain grates, "止まれ" stop signs, and a bus stop.
3. Level crossing and trains: a double-track railway. When a train approaches, the crossing's red lights flash alternately, the bell rings, and the gates lower. A two-car commuter train stops at the station for about 14 seconds and then departs. The windows are transparent, so the seats and hand straps inside are visible.
4. Island-platform station: station name board, platform canopy, benches, and a vending machine.
5. Sakura plaza: a 100-year-old cherry tree with a circular bench around it.
6. Inari shrine: a large vermilion torii plus a row of small torii, stone lanterns, fox statues, a worship hall (copper-green roof, chigi, katsuogi, offering box, hanging bell), a purification fountain (temizuya), jizo statues, ema plaques, a sacred tree, and gravel ground.
7. Riverside levee: two rows of cherry trees forming a blossom tunnel, lanterns, a river, houses on the far bank, and distant mountains.

[How to build the cherry trees (key part)]
- Model them on Somei-Yoshino: the trunk splits low into 3–4 main limbs, which branch recursively three more levels. Branches spread outward and droop slightly at the tips, giving an overall umbrella shape.
- Build the canopy from tens of thousands of "blossom cluster cards": draw five-petal flowers on Canvas (notched petal tips, reddish centers, stamens) with a soft pink base layer behind the flowers. Use alphaTest and double-sided rendering for the cards.
- Add a few pink filler clumps inside the canopy for volume. The outer and upper parts are brighter; the inner and lower parts have warm rose-toned shadows.
- Use pale pink, not neon pink. The canopy sways gently in the wind, a carpet of fallen petals covers the ground under each tree, and petals keep falling through the air (done in a shader).

[Characters]
- 20+ anime-style students and townspeople: walking animation with knee and elbow joints, anime faces drawn on Canvas (large eyes, highlights, blush) that blink, bangs made of separate strands, a variety of hairstyles (long, bob, swinging ponytail, twin tails), sailor uniforms / blazer uniforms / casual clothes. Render characters with two-tone cel shading and a dark outline.
- People walk along the street, chat in the plaza, wait on the platform, pray at the shrine, and ride bicycles along the levee.

[Vehicles]
- Build cars by extruding a side-profile silhouette (with wheel arches, windows, lights, Japanese license plates, and spinning wheels). Cars stop before the level crossing, and while the bell is ringing they wait until the gates rise.

[Lighting and time of day]
- A soft anime-background look: blue sky with white clouds (shader), light haze in the distance, and shadows with a blue-violet tint.
- Switchable between Afternoon / Dusk / Night Sakura. At night, windows, lanterns, and street lamps light up.

[Controls]
- First-person: WASD to walk, Shift to run, Space to jump, F to fly, mouse to look around (pointer lock), number keys to teleport to each location, H to hide the UI, M to mute.
- Mobile: drag on the left half to walk, drag on the right half to look around.
- Collision is enabled, and the player can walk up onto the platform and steps.
- Use Web Audio to generate ambient sound: wind, birdsong, the level-crossing bell, and train running sounds.

make a dynamic 20-second motion graphics and animation video that shows what an incredible motion designer and animator you are, like it's your showreel for a résumé. make it about the cycle of life showing the same individual going from childhood to adolescence to the 9-5 cycle, to family life to elder life and death, then cutting ready to loop to the beginning.  go all out , use whatever you need.

Create a cinematic, highly detailed, interactive 3D rocket launch from an ocean platform at dawn. Deliver the complete project in one HTML file that opens directly in Chrome. Use Three.js + WebGL and procedural assets. Embed assets wherever practical; a reliable CDN may be used for the rendering library.

ART DIRECTION
Create a dramatic transition from a dark, blue-violet ocean before sunrise to warm sunlight above the atmosphere. Use convincing proportions, detailed materials, atmospheric depth, and carefully composed camera angles. The result should feel like a polished miniature spaceflight film.
ROCKET AND LAUNCH PLATFORM
Build a convincing multistage rocket with a shaped nose cone, panel seams, structural rings, interstage connections, engine nozzles, and a payload fairing that separates into two halves.

Create a detailed floating launch platform with a support tower, retracting strongback, service arms, railings, ladders, pipes, equipment, floodlights, and blinking warning beacons. Keep all structures physically connected and correctly positioned.
LAUNCH SEQUENCE
Create an approximately 46-second sequence:

Establishing camera move around the platform.

Service arms and strongback retract.
Engines ignite, illuminating the rocket, platform, and nearby water.
Smoke spreads across the deck as the rocket lifts off and accelerates.
The camera follows the climb from the atmosphere toward space.
The first stage separates and falls away.
The second-stage engine ignites.
The fairing halves separate, revealing a satellite.
The engine shuts down, the satellite deploys, and its solar panels unfold.
Finish with an orbital view of the satellite against Earth’s curved horizon and sunrise.
Compress the flight timeline for presentation while maintaining coherent motion. Avoid sudden position changes, intersecting components, or disconnected effects.
OCEAN, ATMOSPHERE, AND EFFECTS
Use animated shader-driven ocean waves with Fresnel reflections and warm engine-light reflections. Create layered exhaust with a bright core, softer outer flame, and drifting smoke particles.
Smoke should expand, fade, and respond to wind. Exhaust must remain attached to the correct engine through staging. Transition smoothly from atmospheric haze to a dark star field and Earth’s illuminated limb.

CAMERA AND INTERACTION
Use smooth cinematic camera transitions: wide establishing view, low-angle ignition shot, ascent tracking, stage separation, and satellite close-up. Keep the main subject visible in both portrait and landscape layouts.

Include play/pause, replay, and a scrubber with event markers. Seeking must reconstruct the correct rocket configuration, particle state, camera position, and lighting. Replaying must reset the entire sequence cleanly.
Keep the interface minimal and unobtrusive. Allow it to be hidden for recording.

TECHNICAL QUALITY
Use frame-rate-independent animation and efficient particle systems. Reuse geometry and materials where appropriate, handle resizing correctly, and balance visual detail with smooth performance.

Test the final HTML in a desktop browser. Inspect the console and capture screenshots at ignition, liftoff, staging, and satellite deployment. Fix loading errors, clipping, geometry problems, broken seeking, and poor camera framing before delivering the file.

Build your own medieval kingdom.
Your kingdom represents Claude Opus 5.5. Design a magnificent, historically inspired castle that expresses this model’s identity through architecture, heraldry, colors and atmosphere. You have complete artistic freedom within a believable medieval setting.
Do not simply place a logo on a generic castle. Give your kingdom a distinctive architectural character and a coherent visual identity. Invent its coat of arms, royal colors and an original heraldic emblem. Display them on animated banners, shields, gate decorations and the clothing of the castle guards. Place the kingdom’s name above the main entrance.
Create a richly detailed, interactive 3D scene using Three.js and WebGL. Deliver everything in one standalone HTML file that opens directly in Chrome.
THE CASTLE
Build a convincing fortress with a central keep, towers, battlements, curtain walls, an impressive gatehouse, a working drawbridge and a courtyard.
Include carefully modeled stonework, arched windows, wooden doors, roof structures, stairs, balconies, iron fittings and small architectural details. Make the structures believable: towers need interiors or convincing depth, stairs must connect to accessible floors, and bridges must have proper supports.
Surround the castle with an attractive landscape that suits your kingdom: cliffs, hills, a river, a moat, forests or a small village. Design a strong composition that looks beautiful from multiple angles.
LIFE AND INTERACTION
Bring the kingdom to life with guards patrolling the walls, villagers moving through the courtyard, gently waving banners, chimney smoke, birds and flickering lanterns.
Let the viewer:
Open and close the drawbridge and main gate.
Follow a guard on patrol.
Switch between a cinematic overview, the courtyard and the battlements.
Rotate and zoom freely.
Change between daylight, sunset and night.
Keep characters on walkable surfaces. Prevent them from passing through walls, doors or one another.
LIGHTING AND ATMOSPHERE
Create cinematic lighting that reveals the architecture clearly. Use soft shadows, atmospheric depth, convincing water where appropriate and restrained post-processing.
At night, illuminate windows, torches and lanterns while preserving enough visibility to appreciate the castle.
Aim for a sophisticated, finished 3D artwork with distinctive architecture and abundant detail. Avoid a collection of obvious primitive shapes or repetitive towers with no architectural purpose.
TECHNICAL QUALITY
Generate assets procedurally wherever practical. Embed textures and other assets inside the HTML. No local server or build step should be required.
Use instancing and geometry batching where appropriate. Keep animation and camera interaction smooth. Provide a minimal, elegant English interface and a button to hide it.
Actually test the result in desktop Chrome. Capture screenshots, inspect the console, test every interaction and fix rendering errors, floating objects, geometry intersections and camera problems.
Make this a castle people would recognize as YOUR kingdom, even before reading its name.
Return the completed standalone HTML file.

Create an approximately 36-second short film in three.js at 1080×1920 vertical resolution and 30 frames per second. 

【Title】
Breathing Room

【Intended experience】
This film should let viewers experience, without showing any people at all, the feeling of losing time while being driven to stay busy and efficient, followed by the feeling of life filling with rich light the moment they gain some breathing room. At the end, aim to make viewers catch their breath as stained-glass light spreads across the entire floor.

【Setting】
・The interior of a dimly lit stone cathedral. A single pointed-arch window, 15 m high and 10 m wide, is set in the wall at the front.
・Strong light enters from outside the window at a 45-degree diagonal angle, casting a window-shaped patch of light onto the stone floor.
・The window contains stained glass depicting the Virgin Mary in the center, two angels with outstretched wings on either side, and a rose window at the top. Use an original, symmetrical design that does not imitate any existing work.

【Timeline】
0–3 seconds: The light entering through the window is still colorless white light. A soft white shape of light appears on the floor.
3–11 seconds: Gray cubes engraved with words such as “Busy,” “Efficiency,” “Urgent,” and “Deadline” fly in one after another from the foreground, filling the window. Their pace accelerates continuously, and the room grows darker as the window is covered.
11–14 seconds: The window is completely blocked, leaving darkness and silence.
14–19 seconds: A single “Busy” block breaks away from the window, falls, turns into particles of light, and disappears. A shaft of vivid-colored light enters through the opening.
19–26 seconds: Starting from the first opening, the blocks break away in a chain reaction. With each new opening, more columns of multicolored light appear, gradually revealing the stained glass that had been hidden.
26–33 seconds: All the blocks disappear. The camera rises through the columns of light and looks straight down at the floor. Across the entire floor, the Virgin Mary and the angels from the stained glass are projected as vividly colored light.
33–36 seconds: The entire frame is enveloped in dazzling light. The final words appear, and the film ends quietly.

【On-screen words (subtle Mincho-style type, appearing and fading away)】
・“Every day, a little faster.”
・“More efficiently.”
・“Before I knew it, the light had stopped coming in.”
・“I’ll let go of just one thing.”
・“Light enters through the space that opens up.”
・“That light was richer than before.”
・End with the large text: “Abundance lives in the space between.”

【Light and texture】
・Each column of light takes on the color of the stained glass in that location, with airborne dust sparkling in the beams.
・The floor projection must accurately reflect which sections of the window are open; areas occupied by blocks appear in shadow.
・The blocks are an impersonal matte gray, with the words engraved in white.
・The first half is cold and desaturated; the second half uses jewel-like red, blue, and gold. Use this color contrast to convey that the space has become richer.

【Technical requirements】
・Generate the stained-glass design as an image, then calculate the floor light, columns of light, and window display from that same design so they all match.
・Advance time precisely in 1/30-second increments, export one frame at a time, and encode the result as an MP4.

【Final pass】
Render and inspect every scene. Before delivering, fix anything necessary to ensure that the stained-glass design reads as the Virgin Mary and angels on the floor, that the words are legible, and that the motion does not feel abrupt.

three.jsで、縦長1080×1920・30コマ/秒・約36秒の短編映像を作ってください。

【作品名】
余白

【見せたい体験】
効率化に追われて時間を失っていく感覚と、余白を手に入れた瞬間に人生が豊かな光で満たされていく感覚を、人物を一切登場させずに体感させる映像です。最後に床一面に浮かび上がるステンドグラスの光で、見た人が思わず息をのむことを目指します。

【舞台】
・薄暗い石造りの大聖堂の内部。正面の壁に、高さ15m・幅10mの尖頭アーチ型の窓が一つだけある
・窓の外からは、斜め45度の角度で強い光が差し込み、石の床に窓の形の光を落としている
・窓には、中央に聖母、両脇に翼を広げた二人の天使、上部にバラ窓を配したステンドグラスがはまっている。デザインは既存の作品を模さずオリジナルで、左右対称の構図にする

【時間の構成】
0〜3秒：窓から差す光は、まだ色のない白い光。床にやわらかな白い光の形
3〜11秒：「忙しい」「効率化」「至急」「締切」などの言葉が刻まれた灰色の立方体が、部屋の手前から次々と飛んできて窓を埋めていく。飛来のペースはどんどん加速し、窓が埋まるほど部屋は暗くなる
11〜14秒：窓は完全にふさがれ、闇と静寂
14〜19秒：「忙しい」のブロックが一つだけ窓から外れて落ち、光の粒になって消える。空いた穴から、鮮やかな色の一筋の光が差し込む
19〜26秒：最初の穴を中心に、ブロックが連鎖的に外れていく。穴が増えるたびに色とりどりの光の柱が増え、隠れていたステンドグラスが少しずつ姿を現す
26〜33秒：すべてのブロックが消え、カメラは光の柱をくぐり抜けて上昇し、真上から床を見下ろす。床一面に、聖母と天使のステンドグラスが色鮮やかな光として映し出されている
33〜36秒：画面全体がまばゆい光に包まれ、最後の言葉が浮かび、静かに終わる

【言葉（明朝体、控えめに浮かんでは消える）】
・「毎日、もっと速く。」
・「もっと、効率よく。」
・「気づけば、光が入らなくなっていた。」
・「ひとつ、手放してみる。」
・「空いたところから、光が入る。」
・「その光は、前より豊かだった。」
・最後に大きく：「豊かさは余白に宿る」

【光と質感】
・光の柱は、ステンドグラスのその場所の色に染まり、空中の塵がきらめく
・床の投影は、窓のどの穴が空いているかを正確に反映する（ブロックがある場所は影になる）
・ブロックは無機質なマットグレー。言葉は白く刻印されている
・前半は冷たく無彩色、後半は宝石のような赤・青・金。この色の対比で「豊かになった」ことを語る

【技術条件】
・ステンドグラスの絵柄は画像として生成し、床の光・光の柱・窓の表示はすべて同じ絵柄から計算して一致させる
・時間を1/30秒ずつ正確に進めて1コマずつ書き出し、MP4にする

【仕上げ】
各場面を実際に描画して確認し、ステンドグラスの絵柄が床の上で聖母と天使として読み取れるか、言葉が読めるか、動きが唐突でないかを直してから渡してください。

Build a Genshin Impact–style game set in San Francisco.

Create a 4–5 minute cinematic video about the Battle of Austerlitz (1805), built entirely in code.

Research the battle thoroughly and decide for yourself how to tell the story, structure the pacing, explain the strategy, and visualize the events. I want it to be historically accurate, dramatic, easy to understand, and visually exceptional.

Use the attached paintings as visual inspiration, not a strict style requirement. I love their scale, atmosphere, smoke, dramatic skies, cavalry, massed formations, landscape, and sense of chaos. Find a way to translate that feeling into code — but if you can invent a stronger visual language, do it.

Don't make it feel like a generic infographic or strategy game. It should feel like a cinematic historical film that happens to be rendered with code.

You have complete creative control. Surprise me.

build the Eiffel Tower in Three.js.

I want you to imagine a story, that subtly promotes grid genius. Or it can not even have grid genius, but something that aligns with our app and helps us get more eyeballs when posted on social media. Then using threejs/javascript I want you to create full animation, Pixar-level quality out of the story you imagine

Build a real-time 3D game where a pelican cycles through a living seaside world, complete with physics, waves, fishing, dynamic weather, cinematic cameras, autopilot, and adaptive music.

BUILD AN IMPERIAL CITY

build a Bugatti Chiron Super Sport in Three.js.

No 3D model. No textures. No assets.

Create a 30 second animation entirely in code featuring a growth montage, inspired by Kung Fu Panda's training sequence, with the Claude mascot as the main character, showcasing it getting more capable since its initial release across its skillset, such as searching the internet, writing code, creating 3D models, and solving humanity's hardest problems, with emotionally impactful music.

I want you to imagine a story. And then using threejs I want you to create full animation, Pixar-level quality 90s cartoon out of the story you imagine.

Create a seamless looping animation of the water cycle, entirely in code.

Let's develop a game.
The graphics can be either 2D or 3D; choose whichever makes the game systems and overall development easier to analyze and implement.
Personally, I’m imagining a 3D game with a side-scrolling action format.
I’d like to create a stylish atmosphere through the lighting and other elements, so I think 3D could enable more beautiful visuals—for example, streetlamps and lanterns.
The game I want to make is called CatWalk.
The name says it all.
A cat moves from side to side.
A catwalk stretches ahead, and the screen scrolls automatically. The player uses only simple controls, such as jumping, to clear obstacles and gaps while keeping pace with the scrolling speed. In a way, it might have a similar tension and system to Flappy Bird.
However, I want the graphics to feel mature, stylish, and atmospheric.
If possible, I’d love to show the cat moving, running, and jumping with graceful, fluid animation.
You can decide the stage’s setting, but starting with something straightforward, such as a nighttime street, would be fine.
I’d be very happy if a darker setting could make indirect lighting and similar effects look beautiful.
I understand that some things may be possible, impossible, or difficult.
Using my requests as a starting point, please develop something you think you can realistically make.
For now, make one stage playable through a single complete loop.

ゲームを開発しましょう
グラフィックは2D、3Dを問いませんがゲームのシステムなどを分析し開発しやすい方でいいです
個人的には3Dだが画面は横スクロールアクションみたいなものを想定してます
照明などでおしゃれな雰囲気などを出したいので3Dを使うことで美しい表現ができるのではないかと思っています（例：街灯やランタンのような表現など）
作りたいゲームは CatWalk というゲーム
名前の通りです
猫が横に進む
キャットウォークが続いており、画面は自動スクロールするので、その速さに合わせて障害物や穴をジャンプなどのシンプルな操作のみで越えていくスタイル。ある意味でフラッピー的な緊張感とシステムに近いかもしれない。
ただしグラフィックは大人かっこいい、雰囲気ゲーを目指したい
可能であれば猫のしなやかな歩き、走り、ジャンプが表現できればうれしい
ステージの世界観はお任せするが、最初は無難な夜のストリートなどでもいいかと
暗めにすることで間接照明などが美ししく表現できればかなり嬉しい
できることできないこと難しいことがあると思うので
この私の要望をヒントにあなたなりにできそうなものを開発してほしい
まずは１ステージのワンループができるように

build a complete cyberpunk megacity inside Blender with a hero train, procedural architecture, elevated rail systems, rain, volumetric atmosphere, cinematic lighting, multiple camera setups, and a full animated sequence.