Page transitions are the one place a studio site can demonstrate craft without stealing anything, because they happen in the gaps between the things a visitor actually came to see. Done well, nobody registers them as a feature. Done badly, they are a toll charged on every click.

The system uses the View Transitions API with GSAP choreography layered on top. The outgoing page scales back and fades while the incoming one wipes in underneath it, so navigating reads as one continuous surface rather than as a reload. Because the old frame is captured the instant navigation starts, anything still mid-animation would freeze into that snapshot, so interface state is snapped rather than tweened at the moment of the click.

Speed is the real constraint. A transition long enough to admire is long enough to resent by the fifth click, so the timing is tuned to cover the load rather than to sit on top of it. The transition ends roughly when the next page is ready, which makes navigation feel instant even when it is not.

The immersive view gives each project room, with muted loops playing at scale so a visitor can read a piece of work without clicking into it first. Moving through it is meant to feel closer to watching than to navigating.

The recorded list is the opposite by design. Titles, clients, disciplines, and years as plain type, sorted and scannable, with no motion competing for attention. Somebody checking a credit or confirming a delivery date gets what they came for immediately.

Both views read from the same source, so publishing once puts a project in both places with no duplicate entry and no chance of the two drifting out of sync. The switch between them is a state change rather than a page load, which is what keeps it feeling like one site rather than two.

A micro-interaction earns its place by confirming something, not by decorating it. Every interactive element on the site answers the moment it is touched: cursor states that change over media, hover scrubbing on project tiles, buttons that fill rather than blink, counters that actually count.

The discipline is in how small they stay. Each one is short enough that you never wait on it, and none of them announce themselves. If a visitor notices an individual interaction rather than simply feeling that the site is responsive, it has been overbuilt.

On a studio site they also do quiet second duty. They are the smallest available proof that the people who made this care about timing, which is the entire product the studio sells.

Desktop and mobile run different motion budgets rather than the same one at different sizes. Desktop gets the full choreography. Mobile gets a reduced set with fewer simultaneous animations, shorter travel distances, and timings tuned for a thumb instead of a cursor, because a transition that feels considered on a laptop feels sluggish on a phone held at arm's length.

Scroll-triggered reveals are batched so a long case study never asks a phone to animate forty elements at once, and media is served through Cloudflare Stream with real poster frames, so a weak connection shows a still of the work rather than a black rectangle. Autoplaying loops are bounded to what is actually on screen instead of every clip on the page.

The result holds 95+ performance scores across metrics with fast loads globally. On a site whose entire purpose is playing video, that number is not a vanity metric. It is the difference between a reel that starts and a visitor who leaves.

The contact form is built from scratch rather than handed to a third-party form service, because an inquiry is the single most valuable thing that happens on a studio site and it should not depend on someone else's uptime or branding.

A submission does two things at once. It sends a formatted email straight to the studio so a new inquiry surfaces where people already are, and it writes the same inquiry into an internal admin dashboard that is not public, where it can be tracked, assigned, and marked as handled. The email means nothing waits on somebody remembering to log in. The dashboard means nothing gets lost in a thread.

Validation and failure states are handled inline and deliberately, because a form that silently drops a submission is not a bug in a contact form, it is a lost job.

The site also runs a full shopping system for the studio's hats, built directly into the same codebase rather than bolted on through a hosted storefront that would have arrived with its own look, its own checkout, and its own idea of what the brand should feel like.

Behind it is an automated print-on-demand pipeline. An order placed on the site flows straight through to the fulfillment partner, gets produced, and ships without anyone at the studio touching it. No inventory to hold, no packing, no per-order admin.

That automation is what makes merch viable for a small team. It lets the studio run it as a genuine product line instead of a side project that dies the first week everyone is busy on a deadline.

Every decision here comes back to the same tension: prove the studio's capability without ever upstaging the work. The transitions, the motion budgets, and the micro-interactions all resolve that by being felt rather than watched.

The parts that matter most are the ones a visitor never sees. A CRM and a block library mean the site can grow without a developer in the loop. Two views over one dataset mean the work stays correct in both. An in-house form and admin dashboard mean inquiries are tracked rather than trusted to an inbox, and an automated fulfillment pipeline means merch runs itself.

That is the argument for building it from scratch instead of assembling it from services. A studio site is not a brochure that ships once. It is infrastructure the team has to live in, and it is only worth building custom if it makes the next year of work easier rather than harder.