
Higgsfield built a site for creating animations with GPT-6 Astra: draw the first and last frames, generate the in-betweens, and stitch them into a film.
The Hook
“(No spoken dialogue — A high-energy riser leads to a sub-bass drop/impact.)”
Product Reveal
“(No spoken dialogue — Ambient electronic score with pizzicato synths and 808-style percussion.)”
Feature Teaser
“(No spoken dialogue — A riser at 0:06 transitions into the continuing ambient electronic score.)”
Feature Teaser
“(No spoken dialogue — Ambient electronic score continues, building energy.)”
Call to Action
“(No spoken dialogue — A final sub-bass drop/impact at 0:17 concludes the audio track.)”
Synthesizable in Remotion · key components of the ai-creative-workflow-split template.
A vertical split screen divides the view, with the top half showcasing an animated character and the bottom half displaying a user interacting with a tablet UI. Content in both halves animates in sync.
Use a `Split` component with `direction='vertical'`. Animate `opacity` and `transform` properties for content within each pane. Synchronize animations using `useCurrentFrame()` and `interpolate` for precise timing across both sections.
A line art sketch of a character progressively fills with color and detail, transitioning from a raw drawing to a fully rendered animated figure.
Implement a sequence of image overlays or SVG path animations. Use `interpolate` on `opacity` for color layers and `stroke-dashoffset` for line art reveals. Consider `spring` for a more organic fill effect.
As a character animates, faint, semi-transparent previous frames (ghosts) are visible behind the current frame, illustrating the motion path and in-between frames.
Render multiple instances of the character asset with decreasing `opacity` and slight `offset` based on `useCurrentFrame()`. Apply a `blur` or `grayscale` CSS filter to older frames for a distinct ghosting effect.
Purpose. Immediately hook the viewer by showcasing the core AI capability: transforming a simple sketch into a dynamic animation.
Execution. Split screen reveals a hand drawing on a tablet while the top screen shows the sketch coming to life with color and motion.
Purpose. Demonstrate the product's ease of use and the power of AI in generating complex animation sequences.
Execution. Fluid character animation with onion skinning, highlighting the automatic in-between generation within the software UI.
Purpose. Showcase the user's ability to refine and direct the AI-generated content, maintaining artistic vision.
Execution. User interaction with the timeline and drawing tools, making adjustments to the animated character's pose and details.
Purpose. Conclude with a powerful display of the product's output, reinforcing its value proposition.
Execution. A complete, polished animation loop of the character, paired with the Higgsfield branding, leaving a lasting impression.
Objective
A new AI-powered architectural design tool that generates 3D building models from 2D sketches.How the recipe adapts
The split screen would show a user sketching a floor plan on a tablet in the bottom pane, while the top pane dynamically renders a 3D architectural model evolving from the sketch. The 'sketch-to-color' primitive would adapt to 'sketch-to-3D-render', progressively adding textures and lighting. Onion-skinning would show previous iterations of the 3D model as the user refines the sketch, demonstrating the AI's iterative design process. The typography and fast cuts would remain consistent to convey innovation and efficiency.Scene durations (5 scenes, 19.6s)
Key Remotion components

A vibe design partner with an AI-native canvas for turning ideas into UI.

We heard you wanted to use Codex rate limit resets on your own time. Starting today, we’re rolling out the ability to save rate limit resets to use later. We’re starting Go, Plus, Pro, and Business

We just launched something the VFX community has been waiting for. OpenArt VFX lets you apply real visual effects to footage you already have. It lets you mask any region of your video and replace it with AI-generated content that matches the scene's lighting and physics. https://t.co/dnQjrZCWC4