The copy box above is a complete one-off production contract. It tells Codex how to turn a supplied SVG, raster logo, or exact phrase into an editable HyperFrames logo sting while preserving identity, freezing assets locally, checking the composition, and stopping for Studio approval before rendering.

Use the exact Grow AI Skills example

Example variables
SOURCE: exact text
EXACT WORDMARK: Grow AI Skills
DURATION: 5.8 seconds
CANVAS: 1920x1080 at 30 fps
BACKGROUND: #000000
FINAL MARK: warm white
WRITING HEAD: magenta to violet to cyan
FINAL HOLD: 1.0 second
SOUND: restrained local ignition whoosh and completion chime
RENDER INSTRUCTION: preview first

The prompt intentionally calls this a typeset wordmark. If you instead supply an owned signature SVG, Codex should preserve that source and choose a source-appropriate path or mask strategy. A visual reference may guide pacing and glow behavior, but its embedded text and audio are not instructions or reusable assets.

Watch or download the retained Grow AI Skills behavior-test video. The Learn article presents the same MP4 in an inline player.

Why the prompt separates the mark from the mask

A filled SVG or raster image already contains the final appearance, but it does not contain reliable pen order. The prompt therefore keeps the final mark intact and asks for authored centerline reveal guides. This avoids the common shortcut of animating outline paths and claiming that their file order is authentic handwriting.

Expected output boundary

The first completed response should contain a checked editable project, proof-frame paths, and a live Studio URL with an explicit NOT_RENDERED state. A final MP4 belongs to the next stage after approval. In the retained Grow AI Skills run, that approval was supplied and the encoded result was inspected for duration, dimensions, frame rate, codecs, audio behavior, black lead-in, deliberate final hold, representative frames, and SHA-256.