Product Photo Requirements for AI 3D Generation
A factual capture checklist for turning one product photo into a reviewable 3D model, including file limits and difficult materials.
A strong product photo for AI 3D generation does two jobs: it gives the model enough evidence to identify the object, and it gives your team a reliable reference for judging the result.
This guide separates Neyvo3D's actual upload rules from photography recommendations. Recommendations improve the odds of a useful result, but they do not remove the uncertainty of single-view reconstruction.
Neyvo3D's current input requirements
The public Single-View workflow accepts:
- one JPG, PNG, or WebP image;
- a file no larger than 6 MB;
- image dimensions from 128 to 5,000 pixels on each side;
- one product that is fully visible in the frame.
Multi-View upload is not part of the current public workflow. If a guide or old screenshot suggests uploading several angles, do not treat that as a currently available feature.
The recommended composition
Photograph one product, centered, with comfortable space around its outline. Keep every essential part inside the frame. A front or front three-quarter view is often more informative than an extreme top, side, or perspective angle because it shows width and some depth at the same time.
The background should be simple and visually separate from the product. It does not have to be pure white, but the boundary needs to be readable. Avoid hands, lifestyle props, overlapping packaging, text overlays, and other objects that could be mistaken for part of the product.
Google's separate Merchant image guidance recommends a large, full-size product image and says the product should occupy roughly 75–90% of the frame. That is a Google listing recommendation, not a Neyvo3D upload rule, but the same composition is a useful starting point for a clean product source.
Lighting and focus
Use soft, even light. Bright window light through diffusion or a simple light tent can work well. Preserve visible material boundaries without crushing one side into deep shadow.
Focus on the whole product rather than one decorative detail. Motion blur, aggressive portrait-mode blur, heavy JPEG artifacts, and AI background effects remove evidence that the reconstruction process needs.
Difficult materials and shapes
Single-image 3D is less predictable when the image contains:
- clear or translucent glass and plastic;
- mirrors, chrome, or strong specular reflections;
- black objects with lost shadow detail;
- white objects on a white background;
- fur, hair, feathers, or soft fringe;
- very thin wires, spokes, straps, or leaves;
- deep cavities or important rear structures hidden from view;
- severe cropping or lens distortion.
These are physical and visual ambiguities, not merely file-format problems. Single-image reconstruction research can generate a complete mesh from one image, but a complete output is not proof that every hidden surface matches the real object.
A two-minute pre-upload check
Before submitting, ask:
- Is there exactly one intended product?
- Is the full silhouette visible?
- Is the image sharp at the product edges?
- Can you distinguish the product from the background?
- Are the front and some depth cues visible?
- Are reflections or shadows hiding important parts?
- Does the file meet the format, size, and dimension limits?
If the answer to any of the first six questions is no, retaking the photo is usually more useful than enlarging or sharpening the same weak image.
What not to do
Do not upscale a tiny thumbnail and expect new geometric evidence. Do not remove the background so aggressively that thin parts disappear. Do not composite several views into one collage: the current workflow expects one coherent camera view, not multiple cameras placed in one image.
After generation
The input image is only half of quality control. Rotate the result and compare the silhouette, rear, top, base, openings, and materials with the real SKU. If the intended use is an ecommerce listing, have someone who knows the physical product approve the model before publishing.
For an illustrated checklist, use the Neyvo3D photo guide. For the limits behind these recommendations, read Single Image to 3D: Limits and Best-Fit Products.
Sources
Facts checked on 2026-08-19. Platform rules change; verify the linked requirements again before publishing.