Build AI Image Generation Web App

Job Description We are building a consumer-facing web application that transforms a user-uploaded selfie into high-quality themed images. This is a great opportunity for a young developer to build and develop their career. The core product is already built. The focus now is getting the image quality and likeness to a level we can confidently launch, and then continuing to improve and scale the product. This is not a one-off task. We are looking for someone who can own this part of the system long term as the product evolves. This is a great opportunity for a developer that wants to be part of an exciting new development project. Objective Deliver a stable image generation pipeline where: The generated image clearly resembles the uploaded person Outputs are realistic and consistent Background integration is clean and believable The system performs reliably end-to-end Scope of Work Review and improve the existing generation pipeline Improve likeness accuracy (face consistency) Improve image quality (realism, blending, artifact reduction) Ensure consistent outputs across runs Optimize or replace the current AI approach if needed Milestones & Timeline This is a fast-moving project. Timelines matter. Milestone 1: Pipeline Stability + Speed (2–3 days) Goal: Reliable generation with acceptable performance Must deliver: Upload → select → generate → output works end-to-end No failed generations Output renders and downloads correctly Acceptance criteria: 5 consecutive successful generations Each generation completes in ≤ 30 seconds No broken flows or API failures Milestone 2: Likeness + Image Quality (3–5 days) Goal: Generated images clearly resemble the uploaded person Must deliver: Strong face consistency No distorted or “generic face” outputs Clean blending into backgrounds Acceptance criteria: Test set of 5 selfies provided For each selfie: At least 3 acceptable outputs “Acceptable” = clearly recognizable as the same person no visible distortion or warped features natural-looking integration If shown to a third party, they should say: “That’s the same person” — not “kind of similar.” Milestone 3: Backgrounds + Consistency (3–5 days) Goal: Ensure the pipeline works reliably across the full set of provided backgrounds Must deliver: Integration of all provided backgrounds (200+) Consistent generation quality across different background types Proper handling of positioning, scaling, and blending Acceptance criteria: Full background set is usable in the app Test sample across 20–30 backgrounds using multiple selfies No failed generations across tested templates No major drop in likeness or quality No broken selections or UI issues Developer must ensure the system works across all backgrounds without per-template customization. Milestone 4: Launch Readiness (2 days) Goal: Stable, usable application Must deliver: Clean user flow No major bugs Consistent performance Required Experience AI image generation (Stable Diffusion, Flux, etc.) Face consistency techniques (reference conditioning, face swap, inpainting) Experience with APIs/tools such as fal.ai, Replicate, or similar Full-stack capability (Next.js, APIs, backend integration) Ideal Candidate Has built selfie → generated image products Understands tradeoffs between generation vs face swap vs hybrid approaches Can take over and improve an existing pipeline quickly Focused on execution and results To Apply, Answer What approach would you use to improve likeness accuracy? Which models/tools would you use and why? Estimated cost per image? Can you complete Milestone 1 within 2–3 days? Relevant examples (if any) Why This Is Interesting Real product, not a concept Immediate impact on output quality and user experience Opportunity to continue working on the system post-launch You will directly influence how the core AI pipeline evolves Notes Existing application will be shared after initial screening This is a results-driven project with strict milestones Speed and execution matter more than perfect architecture