Owning Orlando

Problem

A self-owned docu-reality series about the power dynamics of Central Florida luxury real estate needed to exist as a fully developed, produceable concept before a single day of shooting could be justified or budgeted — a genre (Owning Manhattan x Succession x Game of Thrones, applied to real Orlando market geography) that if built carelessly around real brokerages and real people would create both defamation exposure and a derivative-title problem against an existing show with a similar name. The project needed a structure that could use real Central Florida market texture for credibility while protecting every actual firm and person from being recognizable as a specific target.

What was built

A complete pre-production package: a locked series bible defining the world (five real Orlando territories mapped to five fictional brokerages, each grounded in actual market dynamics — Winter Park old-money estates, Lake Nona's capital-backed master-plan growth, downtown Orlando's contested high-rise pipeline, Horizon West's new-construction/short-term-rental frontier, and the tourism corridor's hedge-fund vacation-home money), a composite protagonist (Sol Vega, not a real person and not Victor, who stays behind the camera) with a defined arc and fatal flaw, a full 10-episode Season 1 beat sheet, an Episode 1 pilot package ready to prep, a casting-from-scratch plan with sourcing and screen-test protocol, four attorney-review-gated legal document drafts (persona release, location release, on-screen disclaimer, IP/title-clearance memo), a sellable one-sheet, a two-year marketing drip plan, and a dark (unlaunched, noindex) cinematic teaser site with AI-generated video backgrounds built and deployed but deliberately not made public until the pilot exists.

Technical approach

The teaser site is an Astro static build deployed to Cloudflare Pages with a Resend-backed email-capture Function, using seven AI-generated video backgrounds (Higgsfield/Kling) that were ffmpeg-optimized before being committed to the repo rather than served at generation-time resolution. Two infra gotchas were hit and documented rather than silently patched: the Cloudflare API token used for this project has no Zone DNS read/edit scope, so dns_records API calls return success:false — DNS state has to be verified externally via public dig rather than trusted from the API, with actual DNS record changes done manually in the Cloudflare dashboard; and ffmpeg run inside a while read shell loop silently consumes stdin, breaking the loop after the first iteration, unless invoked with -nostdin. A third gotcha: both the Higgsfield and Resend APIs sit behind Cloudflare's edge, and local urllib/curl requests without a browser-like User-Agent header get blocked with a Cloudflare error 1010 rather than reaching the actual API. The Five Houses world-map component (site/src/data/houses.ts) is built and data-complete but deliberately not yet rendered on the live site — staged for a planned mystery-drip marketing reveal rather than shown immediately, which is a real architectural decision (component built, integration withheld) rather than an unfinished feature. The email-capture Function is deliberately two-tier: the Resend contact-creation call is authoritative and any non-duplicate failure returns a 502 to the visitor, while the welcome-email send is best-effort and a failure there only logs a warning rather than blocking a successful signup, so a Resend outage on the email side can't make a working capture look broken. The only bot guard currently in place is a honeypot field that returns a silent 200 when tripped rather than tipping off the bot with an error, with a note left in the source flagging Turnstile or per-IP rate limiting as required before the domain goes fully public. The capture form itself is built progressive-enhancement-first — it posts as a plain HTML form with JavaScript off and is upgraded to a fetch call by a separate enhancement script, so the core signup path doesn't depend on JS executing.

Creative approach

Craft

The visual language is built around AI-generated cinematic video plates rather than stock or licensed footage — city establishing shots, territory plates, and title/transition sequences generated face-free specifically so no real person's likeness appears in marketing before casting exists. The five-territory structure gives the world a literal map (real Orlando geography, invented brokerage names layered on top) that doubles as the franchise's visual identity system, letting each territory carry a distinct color/tone treatment tied to its real market character rather than an arbitrary palette choice — the series' locked grade reserves gold for old-money Winter Park and cool clean whites for capital-backed Lake Nona, with teal-lean shadows and warm skin tones as the season-wide baseline. The visual bible sets one rule above the rest — "the city has power, the people are borrowing it" — and operationalizes it into a fixed three-beat opener for every episode (aerial scale plate, a single cold VO line, then a descent into one human small in a large frame), a coverage test ("could this shot be in Succession? If it could be in a Tuesday-afternoon listing reel, reshoot it"), and an explicit anti-list banning screaming matches, comedic stings, and fourth-wall winks — the same restraint the teaser site's own tone was built to preview.

Reframe

The core creative-legal insight is that the documentary-vs-fiction tension isn't a compromise to manage — it's the load-bearing structure. Real Central Florida geography and real market texture (the actual dynamics of Winter Park old money, Lake Nona's institutional capital, the STR frontier near Disney) provide the credibility a docu-reality format needs, while every brokerage, client, and rival storyline is explicitly invented, which is what makes the show legally producible as a self-owned work about a real industry without exposing any real firm or person as its subject. Protecting the protagonist role as an explicitly composite, castable archetype (not Victor, not any specific real person) rather than a pseudo-anonymized real figure removes the single biggest defamation and release-complexity risk in the format before it becomes a production problem.

Process and what failed

The title itself created a legal gate that had to be resolved before any public marketing: the deliberate genre parallel to Owning Manhattan required a trademark search and a distinguishing subtitle strategy, flagged explicitly in the repo as blocking the public site launch or any paid marketing spend. The teaser site was built and deployed dark (live infrastructure, noindex, no public promotion) as a considered middle state — proving the technical pipeline works end-to-end (video backgrounds, email capture into a verified Resend audience) without committing to a public reveal before casting, budget, and attorney sign-off exist. Five structural pre-production decisions were explicitly deferred to the owner rather than guessed at: who plays the cast, what gear the season shoots on, the Season 1 budget ceiling, the target launch window, and whether post-production capacity runs entirely on the Mac Mini or needs a cloud/second-machine unlock — the project stopped clean at the point where those calls had to be made by a human, not worked around.

Outcome

The creative and legal groundwork is complete — series bible locked, pilot package ready to prep, casting plan and legal drafts written, and a dark teaser site live with a verified end-to-end email-capture pipeline into a real Resend audience. No footage has been shot, no cast is attached, and the trademark/title-clearance question is unresolved. The project's own documentation targets a Fall 2028 launch with real-world casting starting in 2027, contingent on budget and attorney review — this is accurately a fully-developed unreleased concept in pre-production, not a produced or airing show.