Starrd (Independent)Ongoing

AI Video Generator

A native iOS and web AI video generator built on a real production pipeline: a job queue, a credit ledger, and two payment processors, not a prompt box.

Seedance 2.0SupabaseSwiftUI

Concept mockup of Starrd's web generation queue: a dark dashboard with a Starrd sidebar, a "Generation queue" table listing scenes against status and progress — F1 Garage rendering with a half-filled orange progress bar, Kiss Cam and Red Carpet queued with empty bars, K-Pop Idol Cam complete — and a "Job detail" stepper on the right running from Photo uploaded through Queued, Generating, Encoding to Delivered. — Concept mockup — the job queue. A clip takes real time to render, so generation is queued and tracked through discrete stages rather than answered in a single request.

The Problem

Most AI video tools are a prompt box wrapped around a single API call. That breaks down the moment generation takes real time to render, more than one person needs to pay, or the same product has to run on both a phone and a browser.

What I Built

I built and shipped this end to end: product, both clients, backend, and payments. Starrd lets someone pick a scene template, upload one or two photos, and get a cinematic clip back with no prompt engineering required. Native SwiftUI on iOS and a Next.js web app share a single Supabase backend. Generation wraps Seedance 2.0 and runs through a job queue, since a clip takes real time to render rather than returning in one request. A credit ledger tracks usage against credit packs starting at $5.99, one credit per video, sold through Stripe on web and RevenueCat on iOS.

Selected Work

Concept mockup of the Starrd iOS app's scene-template picker: a dark portrait phone screen titled "Starrd", a "Choose a scene" heading, and a two-column grid of template cards showing empty cinematic environments — an F1 pit garage, a floodlit stadium, a spotlit concert stage, a red carpet — labelled F1 Garage, Kiss Cam, K-Pop Idol Cam and Red Carpet, with the F1 Garage card selected in orange and a Continue button above the note "1 credit per video". — Concept mockup — the native SwiftUI scene picker. The user chooses a scene template and uploads a photo; no prompt engineering is asked of them.

Concept mockup of Starrd's billing screen: a dark web app with a Starrd sidebar on Billing, a "Credit ledger" table showing one-credit debits against F1 Garage, Kiss Cam and Red Carpet videos plus a credit-pack purchase row, captioned "1 credit per video", and a "Buy credits" panel reading "Credit packs from $5.99" above an orange Checkout button and the note "Credits never expire". — Concept mockup — the credit ledger. One credit per video, credit packs from $5.99, sold through Stripe on web and RevenueCat on iOS against the same ledger.

How It Works

01Photo upload

02job queued

03Seedance 2.0 generation

04credit ledger debit

05clip delivered to iOS or web client

Outcome

A real production pipeline end to end: a native client and a web client on shared infrastructure, a job queue for long-running generation, a credit ledger, and two payment processors kept in sync. No user, revenue, or download numbers are published.

My role

Solo. Independent product — design, engineering, infrastructure, and launch.

Stack

Next.jsSwiftUI (iOS)SupabaseSeedance 2.0StripeRevenueCatRemotionffmpeg

OneLapse (Independent) · 2024-Present

Back to all work