Claude Code supports GPT Image 2 animated landing pages with mouse-scrub hero sections

• viktoroddy's Claude Code plus GPT Image 2 demo and MengTo's 27-minute tutorial show the same shift: prompt-built landing pages are moving past static mockups into animated, presentation-ready front ends.
• The prompt packs are getting absurdly specific, because viktoroddy's Orbis.Nft prompt spells out mouse-scrub video logic, font loading, Tailwind tokens, and navbar behavior in one reusable spec.
• Creator workflows are splitting across tools, with viktoroddy's post pairing Claude Code with GPT Image 2, while MengTo's Aura thread pitches an AI site builder that can generate and host video inside the product.
• The distribution layer is becoming a product of its own: MotionSites and SceneAI are selling prompt libraries, while MengTo's template post claims 4,700-plus templates that can be converted into DESIGN.md.

MengTo's tutorial post links to a full video walkthrough, MotionSites hosts copy-paste prompt packs, and Aura is framed by MengTo as the only AI site builder in this batch that can generate and host video directly inside the tool. viktoroddy's Axion Studio prompt goes even more granular, down to shader stack order, hover easing curves, and a live London clock, while markproduct's single-prompt demo shows the same aesthetic getting remixed into faster one-shot showcases.

The interesting part is not that Claude Code can make a landing page. That was already table stakes. The new examples lean on motion, richer visuals, and coded interactions that look closer to shipped portfolio work than to a first-pass wireframe.

MengTo's tutorial pitches "animated images 2.0" as a teachable workflow in his post, while markproduct spent three days posting variations like idea to website, AI-designed website, and entire website built with Claude Code. Together they sketch a pattern: creators are using Claude Code less as a novelty generator and more as the implementation layer for polished hero sections.

The strongest evidence in this set is the prompt format itself. Instead of "build me a cool site," these prompts act like compressed creative direction documents with stack choices, interaction logic, visual tokens, and responsive behavior spelled out upfront.

Across the Orbis.Nft spec and the Axion Studio spec, the reusable structure looks like this:

1. Tech stack: React, TypeScript, Tailwind, Vite, plus extras like lucide-react and shaders/react.
2. Visual system: exact hex colors, font families, spacing, shadows, and utility classes.
3. Interaction model: mouse-scrub video seeking, mobile menu state, hover text-scroll effects, arrow rotation, live clock updates.
4. Layout rules: fixed navs, bottom-anchored hero copy, explicit z-index layers, max-width boundaries.
5. Asset instructions: linked video backgrounds, font URLs, inline SVG path data, and component-level dependencies.
6. Responsive behavior: named breakpoints and what changes at each one.

That is why these outputs look less random than older one-prompt website demos. The prompt is doing the work a designer, creative director, and front-end lead would usually distribute across Figma comments, design tokens, and implementation notes.

The motion layer is where this batch separates from the usual AI landing-page churn. viktoroddy's Orbis.Nft prompt describes a fullscreen hero whose video stays paused until the user scrubs left or right, then advances through the timeline by tracking cursor delta and chaining seeked events to avoid dropped seeks.

The Axion Studio prompt pushes the same idea in a different direction with a four-layer shader stack:

• Swirl for the marble base texture
• ChromaFlow for orange blobs that follow the cursor
• FlutedGlass for ribbed distortion
• FilmGrain for analog texture

It also specifies the animation language around the page: duplicated button text that scrolls vertically on hover, arrow icons that rotate -45deg, and a mobile menu that slides up as a bottom sheet. Those are tiny details, but they are exactly the details that make a generated page feel authored.

The ecosystem around these demos is not just models and code. It is prompt inventory. MengTo says Aura now has 4,700-plus templates that can be turned into DESIGN.md, and he linked three live examples for a SaaS automation page, an Atelier AI site, and Elara Footwear.

The other products are packaging the same demand differently. MotionSites is where viktoroddy posted the long-form prompts behind the Orbis.Nft and Axion Studio demos, and SceneAI is the same pitch in lighter form, a one-click bundle of prompts for animated websites. MengTo adds one concrete product claim that did not show up elsewhere in the evidence: Aura can generate and host videos directly in the tool, which points to a tighter path from concept page to live moving asset.