Next.js/Supabase dev to build AI image-generation feature (parallel API calls)

About the product KEYO is an AI content-generation platform. Stack: Next.js 14, Supabase, Clerk, Stripe. A credit system is already in place. This job is to build one self-contained product module ("app") inside the existing platform. Pricing: Fixed-price for the defined scope. Please propose your own price and timeline in your proposal. 1. What the app does A user uploads one photo of a person/character and receives 9 additional shots of the same character from different camera angles (a storyboard), preserving identity — face, outfit, style, and setting. The user then selects the shots they want and downloads them. Reference for the mechanic: higgsfield.ai/apps/shots. Our v1 has no upscale step. v1 category is people/characters only. The architecture should leave room to add other categories (products, scenes) later. Generation model: Nano Banana 2 (Gemini image) with a reference image passed into each generation. This is decided — no alternatives needed. 2. Flow (2 steps) Step 1 — Upload + scene analysis User uploads one image. A vision call (Gemini/GPT) reads the photo and produces a scene description (character, outfit, setting, lighting, style). Based on that description, a set of 9 camera angles for a person is assembled. Note: the description is NOT for consistency. Consistency is held by the reference image passed into every generation. The description exists so the angles fit this specific scene. Step 2 — Generate 9 shots + selection 9 Nano Banana 2 generations, each = reference image + angle prompt. Target resolution: the highest native resolution NB2 supports, aiming for ~2K. No separate upscale step. Generations must run in PARALLEL, not sequentially. The user waits for the slowest shot, not the sum of all nine. This is a hard requirement for both UX and the queue. A 3×3 grid, multi-select of shots, download of selected shots. 3. Ownership split Provided by me (client), NOT in the developer's scope: The finished set of 9 angle prompts for people. Model choice and parameters (Nano Banana 2), API access. UI design / mockups. In the developer's scope: Backend orchestration: vision step → build 9 requests → parallel NB2 calls → assemble results. Generation queue and state handling (in progress / done / error). Grid UI with multi-select and download. Integration into the existing architecture (Next.js 14 / Supabase / Clerk), credit deduction through the existing system. Partial-failure handling. 4. Hard requirements Parallel generations (not sequential), with correct queue management and API rate limits. Partial-failure handling: if 1–2 of the 9 shots fail, do not break the whole grid. Show the ready ones and allow a retry on the failed ones. Job state is persisted (Supabase) so a page reload does not lose the result. Correct credit deduction, with no duplication on repeated clicks (one click = one paid action). 5. What I expect in your proposal Your fixed price for the full scope, and a timeline. Relevant experience with similar work (AI-API integration, queues, parallel generations). Ongoing work This is the first of several planned "apps" for the platform. I'm looking to build a long-term working relationship — strong delivery on this project will lead to continued, ongoing work on future modules.