Indie Hacker Playbooks

Designing Scroll-Driven Websites

Define the visitor journey and emotional curve before building and verifying scroll states

Definition

A scroll-driven website maps scroll position to changes in media, layout, typography or scene state. The interaction can pin sections, scrub video, advance horizontal rails or reveal claims while the visitor controls pace and direction.

Perspectives

Nate Herk (2026-08-22, X)

Start with an interview rather than code: define the scroll journey, first encounter, final belief, available real assets, distinctive behavior, emotional direction and where intensity should peak. In the AI Automation Society redesign, proof appeared before the founder story and community choices, and each claim was paired with a visible source.

Choose one page grammar, one bespoke signature interaction and one engineered emotional peak instead of stacking generic sections or making every section equally dramatic. Preserve brand constraints as hard inputs; use existing photos, product shots, colors and copy before generating missing assets.

Verify the rendered scroll states rather than code alone. Scrollcraft's harness captures screenshots and keyframes to find dead scroll, cues that never become fully visible, frame-dependent contrast failures and video that never decoded, then writes a contact sheet for human review. The first AI Automation Society build took about 30 minutes but still needed a focused feedback pass for pacing, labels, links, copy and a reload bug.

How to apply

  • Fits a brand or product page with strong visual assets and a journey worth pacing; a utility page whose visitor needs immediate scanning may not benefit from scroll choreography.
  • Set the information and proof order before implementation, using Filtering Web Design References to clarify visual direction when no brand system exists.
  • Use Scrollcraft when the stack can support its Node, ffmpeg and browser-verification requirements; keep a simpler implementation path when media cost or performance dominates.
  • Human review remains necessary after automated keyframe checks because technical correctness does not establish meaning, taste or factual accuracy.

Limits

  • Results, build time and quality judgments come from the tool's creator and one featured redesign, not a conversion or usability test.
  • The source reports one remaining reload bug and several factual or interaction errors after the first build.
  • Generated media adds cost and scroll-linked effects can create mobile performance, accessibility and motion-sensitivity risks not measured here.

Original link

On this page