PortfolioCase study
CP-01Live
Cinematic Portfolio
A portfolio, on film.
A scroll-driven film portfolio — AI-generated footage of Tushar as the backdrop, scrubbed like an Apple product page, with no build step.
0 ms
frames over 16 ms in the scrub benchmark
Measuredas of 26 Aug 2026
mean 0.04 ms per frame
31 / 31
QA checks passed across three passes
Measuredas of 26 Aug 2026
8/8, 11/11, 12/12
Evidence keyMeasured
The question
Recruiters ask one thing first: who is Tushar Pathak?
A list of links answers what he did, not who he is. The bet was a short film that answers both as you scroll.
The film
Scroll, and the film plays with you.
Three scenes — an opening, the work, a closing walk — scrub frame by frame with the page. Under reduced motion, the film becomes still frames.
Decisions
Cinematic, but never at the reader’s expense.
- Could have
- A conventional page of cards
- Chose
- A film as the backdrop
- Because
- The PRD starts from the recruiter’s first question: who, not what.
- Could have
- Motion for everyone
- Chose
- A static fallback under reduced motion
- Because
- The film must never be the price of reading the page.
- Could have
- A framework and a build step
- Chose
- Static HTML, CSS and vanilla JS
- Because
- Nothing to build; the film and the page ship as plain files.
How it was made
Generated, cut, then scrubbed.
- Generate the filmHiggsfield, credits pre-flighted
- Post-processffmpeg and headless Chrome
- 289 scrub frames
- Smooth scrollLenis moves the frames
- Static fallbackunder reduced motion
What I learned
The tooling is part of the craft.
Know the tool’s gates
Plan gating, refunds on a failed render and start-image versus reference behaviour shaped the film.
Pre-flight the spend
Estimating credits before rendering meant the film cost exactly what was planned.
Sources
Everything here is backed by a real artifact.
- PRD
- Build ledger
- Design