AI Product Image Generation Specialist for 3D Model Input Photos

Project Overview Bond Studio makes software that lets users capture their space with their phone and see a visualization of how that room could look if it were remodeled. Users can browse products, select them, and see them visualized in their space. We are improving the image inputs used for AI-assisted 3D model generation. When we ingest a product, we need a clean, canonical product image that can be sent into our 3D generation workflow. Where retailer images often include clutter, other products, bad angles, inconsistent lighting, or missing or unclear product details our input images must be clean, focused on the product, and at a view angle that makes all details of the product clear for the asset generator. This is not a generic marketing-image or chatbot project. We need someone who can use modern image generation and editing tools to preserve product identity, geometry, finish, details, and attributes while normalizing the image into a consistent model-input photo. The end process must be fully automated, consistent, fast, cost-effective, and accurate. Scope of Work You will help us design and test a repeatable image generation and editing workflow for product photos. The goal is to transform imperfect source imagery into clean, trustworthy inputs for downstream 3D model generation. The work will include: Reviewing product source images and metadata to understand which details must be preserved. Designing prompts, masks, control images, reference-image workflows, or other generation steps that produce consistent product-faithful outputs. Generating or editing images into clean product-only model-input photos. Removing scene clutter, backgrounds,and other unwanted details. Normalizing camera view, perspective, lighting, and background. Preserving or verifying important product details. Producing a repeatable prompt or workflow recipe that can be executed in a fully automated way. Helping define a practical quality gate for when an image is safe to send into 3D model generation. Expected Deliverables A workflow for generating or editing these images, including tools, prompts, parameters, and QA checks. A small batch of before-and-after examples that meet the requirements. Technical Expectations You may use any image generation or editing stack that can produce reliable product-faithful outputs: Flux, Stable Diffusion, ComfyUI, Photoshop or Firefly, Runway, Ideogram, LLM-based image models, segmentation tools such as SAM, or other tools. We care more about repeatability and quality than a specific tool choice. You should be able to explain how your workflow prevents common failures: changed geometry, altered material or finish, hallucinated hardware, missing details, bad perspective, inconsistent lighting, and artifacts where items or backgrounds were removed. Qualifications Strong portfolio of AI image generation, image-to-image editing, inpainting, or product-image cleanup. Experience preserving real product identity and geometry while editing backgrounds, camera angle, lighting, and unwanted objects. Comfort working with interior-design, furniture, e-commerce, retail, or architectural product imagery. Strong visual QA judgment: perspective, lighting, materials, product details, and plausible commercial product photography. Experience with segmentation, masks, control images, depth or normal guidance, or reference-image conditioning is a plus. Experience with 3D asset pipelines, image-to-3D workflows, or product visualization is a plus. Ability to document a repeatable process clearly enough for our team to evaluate. Clear written communication and willingness to show before-and-after examples and explain tradeoffs.