# From Product image to 3D Model: No-Scanner Workflow > Source: https://blog.pixlnexs.com/product-image-to-3d-model-no-scanner-workflow/ > Published: 2026-09-10 · Author: kishore > By the Pixlnexs Studio Team. Pixlnexs develops interactive 3D and AR product visualization for eCommerce brands, and this photo-to-3D workflow is the --- > **Quick answer:** Converting product image to 3D model uses photogrammetry or AI-assisted reconstruction to build a mesh from 2D images, without needing a dedicated hardware scanner. The workflow involves capturing 30–100 high-resolution images from multiple angles, processing them into a textured 3D asset, cleaning up the mesh and optimizing the result for the web. It suits e-commerce brands that want interactive 3D views without investing in specialized scanning equipment and the full process typically runs from a couple of days to about a week depending on product complexity and required detail. By Bali Balaji, Pixlnexs Studio Team. Pixlnexs develops interactive 3D and AR product visualization for eCommerce brands and this image-to-3D workflow is the one our production team runs day to day. Table of Contents [Toggle](#) - [Key Takeaways](#Key_Takeaways) - [Why Skip the Scanner? The Economics of image-Based 3D](#Why_Skip_the_Scanner_The_Economics_of_image-Based_3D) - [The Operator-Grade Workflow: 8 Steps to a Production-Ready Asset](#The_Operator-Grade_Workflow_8_Steps_to_a_Production-Ready_Asset) [Step 1: Shoot List, Angle Count and Lighting Setup](#Step_1_Shoot_List_Angle_Count_and_Lighting_Setup) - [Step 2: Background Removal and Masking](#Step_2_Background_Removal_and_Masking) - [Step 3: Photogrammetry or AI Mesh Reconstruction](#Step_3_Photogrammetry_or_AI_Mesh_Reconstruction) - [Step 4: Retopology and Cleanup](#Step_4_Retopology_and_Cleanup) - [Step 5: UV Unwrap and Texture Bake From Photos](#Step_5_UV_Unwrap_and_Texture_Bake_From_Photos) - [Step 6: Material / PBR Pass](#Step_6_Material_PBR_Pass) - [Step 7: Optimization and Export to GLB](#Step_7_Optimization_and_Export_to_GLB) - [Step 8: QA in a Web Viewer](#Step_8_QA_in_a_Web_Viewer) - [Workflow Comparison: Photos vs. Scanner vs. CAD](#Workflow_Comparison_Photos_vs_Scanner_vs_CAD) [Which Method Should You Choose?](#Which_Method_Should_You_Choose) - [Common Pitfalls When Converting image to 3D Models](#Common_Pitfalls_When_Converting_image_to_3D_Models) [Poor Lighting in Source Photos](#Poor_Lighting_in_Source_Photos) - [Insufficient Photo Overlap](#Insufficient_Photo_Overlap) - [Reflective or Transparent Surfaces](#Reflective_or_Transparent_Surfaces) - [Ignoring the Background](#Ignoring_the_Background) - [Skipping Mesh Cleanup](#Skipping_Mesh_Cleanup) - [Integrating the Finished Model Into Your Store](#Integrating_the_Finished_Model_Into_Your_Store) [Choosing a Web Viewer](#Choosing_a_Web_Viewer) - [Mobile Compatibility](#Mobile_Compatibility) - [Providing a Fallback](#Providing_a_Fallback) - [Measuring Whether It Is Working](#Measuring_Whether_It_Is_Working) - [Scaling the Workflow Across a Catalog](#Scaling_the_Workflow_Across_a_Catalog) [Standardize the Capture Process First](#Standardize_the_Capture_Process_First) - [Batch by Product Similarity](#Batch_by_Product_Similarity) - [Build a QA Checklist, Not Just a QA Step](#Build_a_QA_Checklist_Not_Just_a_QA_Step) - [Decide Where Automation Helps and Where It Does Not](#Decide_Where_Automation_Helps_and_Where_It_Does_Not) - [Cost Considerations for the No-Scanner Workflow](#Cost_Considerations_for_the_No-Scanner_Workflow) - [When This Workflow Is Not the Right Fit](#When_This_Workflow_Is_Not_the_Right_Fit) - [A Note on Consistency Across a Growing Catalog](#A_Note_on_Consistency_Across_a_Growing_Catalog) - [Ready to Turn Your Product image Into 3D Assets?](#Ready_to_Turn_Your_Product_image_Into_3D_Assets) [Have product photos but no scanning hardware?](#Have_product_photos_but_no_scanning_hardware) - [Frequently Asked Questions](#Frequently_Asked_Questions) [How many images do I need to convert product images to a 3D model?](#How_many_images_do_I_need_to_convert_product_images_to_a_3D_model) - [Can I use a smartphone to take photos for 3D modeling?](#Can_I_use_a_smartphone_to_take_photos_for_3D_modeling) - [What is the best software for converting image to 3D models?](#What_is_the_best_software_for_converting_image_to_3D_models) - [How long does it take to convert product image to 3D model?](#How_long_does_it_take_to_convert_product_image_to_3D_model) - [Can I use these 3D models for augmented reality experiences?](#Can_I_use_these_3D_models_for_augmented_reality_experiences) - [What is the difference between a 3D scan and a 3D model built from image?](#What_is_the_difference_between_a_3D_scan_and_a_3D_model_built_from_image) - [How do I fix holes in my 3D model after photogrammetry?](#How_do_I_fix_holes_in_my_3D_model_after_photogrammetry) - [Is it worth investing in this workflow for a small catalog?](#Is_it_worth_investing_in_this_workflow_for_a_small_catalog) - [Can I reuse the same photos I already have for product listings?](#Can_I_reuse_the_same_photos_I_already_have_for_product_listings) - [What is a reasonable file size target for the finished GLB?](#What_is_a_reasonable_file_size_target_for_the_finished_GLB) - [Do I need a different workflow for products with moving or hinged parts?](#Do_I_need_a_different_workflow_for_products_with_moving_or_hinged_parts) - [Can I outsource just part of this workflow, like cleanup and optimization and keep photography in-house?](#Can_I_outsource_just_part_of_this_workflow_like_cleanup_and_optimization_and_keep_photography_in-house) ### Key Takeaways - **Hardware-agnostic:** a smartphone, DSLR or mirrorless camera is enough no dedicated scanner required. - **Image quantity matters:** aim for 30–50 images for a standard product; complex items may need 100 or more. - **Lighting is critical:** consistent, diffuse lighting keeps shadows from being misread as geometry. - **Post-processing is mandatory:** raw photogrammetry output is high-poly and noisy retopology and texture baking are essential before web use. - **Cost efficiency:** the primary cost is time and process discipline rather than equipment, which makes the workflow scalable across a catalog. ## Why Skip the Scanner? The Economics of image-Based 3D For most e-commerce brands, the barrier to [3D product visualization](https://blog.pixlnexs.com/best-3d-product-visualization-platforms/) is not technical skill but capital expenditure. Dedicated 3D scanners structured light or laser triangulation devices carry a significant hardware cost and often require controlled environmental conditions, such as consistent lighting and sometimes a turntable rig, which adds operational overhead on top of the purchase price. Converting product image into a 3D model using a smartphone or a DSLR removes that barrier. The “[photogrammetry](https://blog.pixlnexs.com/scan-a-product-with-phone-camera/)-lite” approach works because most manufactured products have simpler, more predictable geometry than organic subjects like faces. By capturing a product from many angles, software can triangulate matching surface points across images and reconstruct a mesh. This works especially well for rigid, matte-finished items. It is not a magic bullet, though. Output quality tracks input quality directly. A blurry photo or a poorly lit angle produces a noisy mesh or texture artifacts. The workflow depends more on capture discipline than on which software you choose. For background on the underlying technique, see the [Wikipedia overview of photogrammetry](https://en.wikipedia.org/wiki/Photogrammetry). ## The Operator-Grade Workflow: 8 Steps to a Production-Ready Asset ![The Operator-Grade Workflow 8 Steps to a Production-Ready Asset](https://blog.pixlnexs.com/wp-content/uploads/2026/09/The-Operator-Grade-Workflow-8-Steps-to-a-Production-Ready-Asset-1024x683.png) This is the pipeline our production team uses to turn raw image into a web-ready 3D asset, refined across many projects with real-world products rather than idealized test objects. ### Step 1: Shoot List, Angle Count and Lighting Setup Before picking up the camera, plan the capture. The goal is enough overlapping images for the software to match features reliably. For a standard product like a bottle or box, 30–50 images is usually enough; for intricate items like a watch or jewelry, 100 or more may be needed. **A practical angle pattern:** 1. **Top-down pass** to capture the top surface. 2. **Bottom-up pass** to capture the base, where relevant. 3. **Side passes**, rotating the product in small increments around the vertical axis. 4. **Elevation passes** from low, mid and high camera angles to cover curvature. **Lighting setup:** * Use diffuse lighting direct sunlight or harsh studio lights create specular highlights that confuse photogrammetry software. * If using a turntable, make sure it is stable; vibration causes motion blur. * Keep the background neutral (white or gray) and free of patterns that could be mistaken for part of the product. A simple light tent is often enough to provide the diffuse, shadow-free environment the capture needs. Trying to fix uneven lighting in post-processing is rarely worth the time compared to getting it right during the shoot. ### Step 2: Background Removal and Masking Most photogrammetry software can handle backgrounds automatically but manual masking gives more control, especially if the background is not perfectly uniform. **Why mask:** * **Accuracy** prevents the software from interpreting shadows or reflections as geometry. * **Texture quality** keeps the texture map limited to the product itself. * **File size** smaller, cleaner textures load faster. **Tools:** Photoshop’s Select Subject or Pen tool for precision, remove.bg for quick automated removal (checking edges carefully) or GIMP as a free alternative. Keep the original image with its background on a separate layer for reference. A mask that is too tight can cut off part of the product; one that is too loose lets background noise into the texture. A slight feather on the mask edge avoids hard lines in the final texture. A poorly masked image is a common cause of a visible “halo” artifact around a product in the finished model worth catching at this stage rather than after processing. ### Step 3: Photogrammetry or AI Mesh Reconstruction This is the core of the workflow: feeding the images into software that analyzes overlapping features to build a 3D mesh. **Photogrammetry software:** RealityCapture (fast, accurate, handles large datasets well) Meshroom (free and open-source, good for getting started) and Agisoft Metashape (a common industry choice with a good balance of speed and quality). **AI-assisted reconstruction:** tools using neural radiance fields (NeRFs) or mobile apps combining LiDAR and photogrammetry, can process images faster but may be less accurate on reflective or transparent surfaces and harder to edit afterward. For production e-commerce use, photogrammetry software generally gives more reliable, controllable results. AI tools are useful for quick previews but tend to struggle with glossy or transparent surfaces. **Processing tips:** use the highest-resolution images available rather than downscaling beforehand, set mesh density to “high” for the initial reconstruction (you can decimate later) and check the alignment report a failure to align some images usually means insufficient overlap or a lighting change mid-shoot. ### Step 4: Retopology and Cleanup The raw mesh from photogrammetry is high-poly, sometimes containing millions of triangles far too heavy for web use. Reducing polygon count while preserving shape is essential for performance, easier editing and cleaner UV maps. **Tools:** Blender (free, with a Decimate modifier and Remesh tool) ZBrush (a common choice for sculpting and retopology) and 3ds Max (with Quad Draw for manual retopology). **Process:** import the raw mesh, remove obvious artifacts such as floating triangles or holes, decimate the polygon count toward a web-appropriate range (commonly in the tens of thousands of triangles for a standard product) verify the silhouette still reads accurately from all angles and smooth the normals so lighting looks correct. Do not decimate too aggressively over-reducing polygon count loses texture-relevant detail. A two-pass approach decimate partway, then manually clean up remaining artifacts tends to give a cleaner result than one aggressive pass. ### Step 5: UV Unwrap and Texture Bake From Photos UV unwrapping flattens the 3D mesh into a 2D texture map, which is where the original photos get used to build the final texture. **Tools:** Blender’s Smart UV Project or Unwrap tools, Substance 3D Painter for texturing and Marmoset Toolbag for baking and preview. **Process:** select the mesh, unwrap it into a 2D layout, arrange UV islands to make efficient use of texture space and bake the texture from the high-poly mesh onto the low-poly mesh to transfer detail. **Baking tips:** use a high-resolution texture map appropriate to the product’s level of detail, bake diffuse, normal and roughness maps as the essential trio for PBR materials and check for stretching, which shows up as visibly distorted texture in the final render. ### Step 6: Material / PBR Pass PBR (Physically Based Rendering) materials simulate how light interacts with a surface, which is essential for a realistic result. **Key PBR maps:** albedo/diffuse (base color) normal (surface bumps and detail) roughness (how matte or glossy the surface is) metallic (how metallic the surface behaves) and ambient occlusion (shadowing in crevices). **Tools:** Substance 3D Painter, Blender or Marmoset Toolbag. Work from a reference photo of the real product while texturing rather than guessing at color or finish. The roughness map deserves particular attention for product visualization it determines how light reflects off the surface and getting it wrong is one of the most common reasons a 3D product model looks slightly “off” compared to the real item. ### Step 7: Optimization and Export to GLB GLB (Binary glTF) is the standard format for web-delivered 3D, compact, efficient and supported by most web viewers, maintained as an open specification by the [Khronos Group](https://www.khronos.org/gltf/). For a full comparison against other formats, see [GLB vs USDZ vs OBJ: 3D File Formats Explained](https://blog.pixlnexs.com/glb-vs-usdz-vs-obj-3d-file-formats-explained/). **Optimization tips:** compress textures (JPEG or WebP rather than large PNGs for most maps) enable Draco compression to reduce mesh size substantially with minimal visible quality loss and remove any unused materials, since they add file size without adding value. **Tools:** Blender’s glTF exporter with Draco enabled, `gltfpack` for command-line optimization and MeshLab for additional mesh cleanup. **Process:** export from your 3D software as glTF/GLB, compress with a tool like `gltfpack`, check the resulting file size against your performance budget and test the model in an actual web viewer before considering it final. ### Step 8: QA in a Web Viewer The final step is testing the model in a real-world web viewer, which is where issues invisible in your 3D software tend to surface. **What to check:** load time, visual quality (artifacts or distortion) interactivity (smooth rotate and zoom) and mobile performance specifically, since a large share of shoppers will view the model on a phone. **Tools:** Three.js, Babylon.js or the `` web component. Testing on mobile devices, across a range of hardware rather than just your own newest phone, catches performance problems before they reach shoppers. ## Workflow Comparison: Photos vs. Scanner vs. CAD Method | Relative Cost | Typical Turnaround | Fidelity | Best For | Photogrammetry (photos) | Low to medium | A few days | High, with good lighting discipline | Complex organic shapes, one-off products, budget-conscious teams | Dedicated 3D scanner | High (equipment) | Fast per unit once set up | Very high | High-volume production, intricate detail, consistent batch capture | Manual CAD modeling | Medium to high | Longer, artist-dependent | Variable, depends on the artist | Simple geometric shapes, engineering precision, stylized assets | ### Which Method Should You Choose? - **Choose photogrammetry** if you have a complex organic-ish shape (a shoe, a handbag, a food item) and want a realistic model without the cost of a dedicated scanner the most accessible option for small and mid-size teams. - **Choose a dedicated scanner** if you are producing a large, ongoing batch of similar products and need consistent, repeatable capture the equipment cost amortizes over volume. - **Choose manual CAD** if your product has simple, repeatable geometry (a bottle, a box, furniture with straight lines) and you need precise dimensions for manufacturing or engineering reference, not just visualization. In practice, a hybrid approach is common: photogrammetry captures the base shape quickly and manual CAD adjustments refine specific dimensions or details where precision matters more than photographic realism. ## Common Pitfalls When Converting image to 3D Models ![Cost Considerations for the No-Scanner Workflow](https://blog.pixlnexs.com/wp-content/uploads/2026/09/Cost-Considerations-for-the-No-Scanner-Workflow-1024x512.png) ### Poor Lighting in Source Photos Shadows, reflections and uneven illumination confuse photogrammetry software and cause artifacts. Use a lightbox or softbox for even, diffused lighting, avoid direct sunlight or harsh overhead lights and shoot in a controlled space rather than a room with mixed light sources. ### Insufficient Photo Overlap Photogrammetry relies on overlapping features between photos to reconstruct shape. Too little overlap leaves holes or distortions. Aim for substantial overlap between consecutive photos, move in a consistent circular path at a steady distance and shoot more photos than you think you need. ### Reflective or Transparent Surfaces Glossy, metallic or transparent surfaces are difficult for photogrammetry because they reflect the environment instead of showing their own texture. A temporary matte spray or powder, a polarizing filter on the lens or switching to manual CAD modeling for genuinely transparent products can all help. ### Ignoring the Background A busy or cluttered background can introduce unwanted geometry into the reconstruction. Use a plain, neutral background, remove unrelated props or key out a green-screen background in post-processing if you cannot control the physical set. ### Skipping Mesh Cleanup Raw photogrammetry output is rarely ready to use as-is. Skipping cleanup leaves noise, holes and unnecessary geometry that make texturing and rendering harder. Budget real time for this step rather than treating it as optional. ## Integrating the Finished Model Into Your Store ### Choosing a Web Viewer Options range from Google’s `` web component, which is straightforward to integrate, to Three.js or Babylon.js, which need more setup but offer more control over the interaction design. ### Mobile Compatibility Test on a range of mobile browsers and devices, optimize the interaction for touch (pinch-to-zoom, swipe-to-rotate) and confirm the model loads at an acceptable speed on typical mobile networks. ### Providing a Fallback Not every browser fully supports WebGL and some users disable it. Detect support and fall back to a static image or a short video where the 3D viewer cannot run, so no shopper hits a broken page. ### Measuring Whether It Is Working Track how shoppers interact with the model and compare the conversion rate and return rate of products with a 3D view against comparable products without one over a long enough window to smooth out seasonal noise, being careful to separate the effect of the 3D model from unrelated changes running at the same time. ## Scaling the Workflow Across a Catalog Converting one product from image to a 3D model is a manageable project. Converting a hundred-SKU catalog is a different kind of problem and it is worth planning for before you start. ### Standardize the Capture Process First Before scaling up, lock down a repeatable capture setup: the same lighting rig, the same camera distance and angle pattern, the same background. Consistency at capture time saves far more time downstream than any software shortcut, because every product that deviates from the standard setup needs individualized troubleshooting later. ### Batch by Product Similarity Group products by shape and material similarity all similarly shaped bottles together, all similarly shaped boxes together so that retopology and texturing settings developed for one item transfer with minimal adjustment to the next. Treating every product as a one-off resets your learning curve each time. ### Build a QA Checklist, Not Just a QA Step At catalog scale, an informal “does it look right” pass does not scale. A written checklist silhouette accuracy from four angles, texture seam check, file size against budget, load test on a reference mobile device keeps quality consistent even as different team members handle different batches. ### Decide Where Automation Helps and Where It Does Not Steps like texture compression and GLB export can be scripted and batched reliably. Steps like judging whether a roughness map correctly conveys a specific material’s real-world finish still benefit from a human check, at least for the first pass on any new product category. Automating too early on judgment-dependent steps tends to produce a catalog of technically valid but visually inconsistent models. ## Cost Considerations for the No-Scanner Workflow ![Common Pitfalls When Converting image to 3D Models](https://blog.pixlnexs.com/wp-content/uploads/2026/09/Common-Pitfalls-When-Converting-image-to-3D-Models-1024x683.png) The dominant cost in this workflow is people’s time, not equipment, which changes how you should think about budgeting compared to a scanner-based approach. - **Photography time** scales with the number of angles captured and how controlled your existing photo setup already is a team already doing studio product photography has a head start. - **Reconstruction and cleanup time** scales with product complexity far more than with the software license cost; a simple bottle takes a fraction of the cleanup time a detailed piece of jewelry does. - **Texturing time** depends on how many distinct materials the product has a single-material item (a plain ceramic mug) is quick, while a product combining metal, fabric and plastic takes longer to get each material looking right. - **QA and optimization time** is relatively fixed per product once your pipeline and checklist are established, which is part of why standardizing the process early pays off at scale. Because the main cost driver is labor and process maturity rather than a large fixed hardware purchase, this workflow tends to have a lower barrier to entry than a scanner-based approach but a less predictable per-unit cost until your team has been through the process enough times to have a reliable sense of how long each product type actually takes. ## When This Workflow Is Not the Right Fit The no-scanner workflow is versatile but it is not the correct tool for every situation. - **Highly reflective or fully transparent products** (polished metal, clear glass, mirrored surfaces) are the hardest case for photogrammetry, since the technique depends on the software recognizing consistent surface features across photos and a mirror-like surface reflects the environment instead of showing itself. A dedicated scanner with structured light or a manual CAD-plus-reference-photo approach, is often more reliable here. - **Products requiring exact manufacturing tolerances**, such as parts destined for CAD-based engineering review, are better served by manual modeling from technical drawings or a scanner explicitly built for metrology, since photogrammetry accuracy is generally sufficient for visualization but not guaranteed to hold to tight engineering tolerances. - **Very large batches on a tight deadline** may favor a dedicated scanning rig despite the upfront hardware cost, since a scanner with a fixed capture routine can process a large volume of similar products faster per unit than a photography-plus-manual-cleanup pipeline, once the equipment is already in place. - **Products that change size or shape slightly between units** (soft goods, some food items) introduce a mismatch between the captured reference unit and what actually ships, which is a modeling-fidelity problem no capture method fully solves it is worth deciding upfront how much of that natural variation you are comfortable representing with a single model. Recognizing which category a product falls into before starting the capture saves time compared to discovering the limitation partway through processing. ## A Note on Consistency Across a Growing Catalog As a catalog grows, small inconsistencies in the capture-to-web pipeline tend to compound. A texture resolution that seemed fine for the first ten products can look noticeably out of step with newer products captured under an improved lighting setup. Two practical habits help keep a catalog visually consistent over time: - **Document your current pipeline settings** resolution, compression targets, lighting setup as a living reference and update older assets in batches when the reference changes materially, rather than letting the catalog drift into an unplanned mix of quality levels. - **Periodically spot-check older assets** against current standards, particularly before a major site redesign or a push into a new sales channel, since a visible quality gap between older and newer 3D assets is more noticeable to shoppers than a similar gap in ordinary product photography. ## Ready to Turn Your Product image Into 3D Assets? If you already have decent product photography and want to add interactive 3D without investing in scanning hardware, this is exactly the workflow our production team runs for e-commerce catalogs. We handle capture guidance, reconstruction, cleanup, texturing and web optimization end to end. ### Have product photos but no scanning hardware? Pixlnexs runs this exact photogrammetry pipeline end to end capture guidance, reconstruction, cleanup, texturing and web optimization. [Talk to Pixlnexs](https://pixlnexs.com/3d-product-modeling/) ## Frequently Asked Questions ### How many images do I need to convert product images to a 3D model? There is no fixed number but a reasonable starting point is 30–50 photos for a standard product, with more needed for complex or highly detailed items. The priority is sufficient overlap and full coverage from all relevant angles rather than hitting an exact count. ### Can I use a smartphone to take photos for 3D modeling? Yes, though a DSLR or mirrorless camera generally gives better results. If using a smartphone, a tripod and a remote shutter (or timer) help avoid camera shake and consistent, diffused lighting matters more than the camera itself. ### What is the best software for converting image to 3D models? There is no single best option it depends on budget and workflow. RealityCapture, Metashape and Meshroom are common choices for the reconstruction step, with Blender frequently used for the cleanup and texturing stages that follow. ### How long does it take to convert product image to 3D model? It depends on product complexity and the quality of the source photos. A standard product often takes a few days from capture to a web-ready asset; more complex or detailed products can take longer, mainly due to additional cleanup and texturing time. ### Can I use these 3D models for augmented reality experiences? Yes, provided the model is optimized for mobile and exported in a compatible format GLB for Android AR, USDZ for [iOS AR Quick Look](https://blog.pixlnexs.com/usdz-files-ar-quick-look/). Test the AR experience on actual devices rather than relying on a desktop preview. ### What is the difference between a 3D scan and a 3D model built from image? Both are ways of capturing a real object digitally. A dedicated scanner captures geometry directly using structured light or laser triangulation, generally with less manual cleanup. Photogrammetry reconstructs geometry indirectly from overlapping photos and typically needs a more involved cleanup and texturing pass but requires no specialized hardware. ### How do I fix holes in my 3D model after photogrammetry? Holes usually trace back to insufficient photo overlap or inconsistent lighting during capture. The most reliable fix is capturing additional photos of the problem area and reprocessing; for isolated small holes, manual mesh repair in a tool like Blender can also work. ### Is it worth investing in this workflow for a small catalog? It can be, particularly for products that benefit from a detailed visual representation or that see high return rates tied to appearance mismatches. A practical approach is to start with a handful of representative products, measure the impact and expand from there rather than converting an entire catalog at once. ### Can I reuse the same photos I already have for product listings? Sometimes partially but usually not fully. Standard e-commerce product photography is typically optimized for a small number of hero angles rather than the dense, overlapping angular coverage photogrammetry needs. Existing photos can serve as a useful reference for color and material accuracy during texturing but a dedicated capture pass built around the angle count and lighting consistency described earlier in this article will generally give a more reliable reconstruction. ### What is a reasonable file size target for the finished GLB? There is no universal number, since it depends on product complexity and your performance budget but a common target for a standard e-commerce product is a file small enough to load comfortably on a typical mobile connection without a long wait often a few megabytes after Draco and texture compression. Test against your own analytics for typical visitor connection speed rather than assuming a single target fits every catalog. ### Do I need a different workflow for products with moving or hinged parts? Yes, at least for the animation layer. The capture and base-model steps in this workflow are the same but a product with a hinge, a lid or an adjustable component needs its mesh segmented into the moving parts during retopology, with rigging and animation added afterward a step beyond the static-model workflow covered here, closer to what a configurator’s asset pipeline requires. If that describes more of your catalog than a handful of exceptions, it is worth reading about configurator-specific asset pipelines separately, since the segmentation and rigging work changes the scope of the project meaningfully compared to a purely static model. ### Can I outsource just part of this workflow, like cleanup and optimization and keep photography in-house? Yes and it is a common split for teams that already have a working product photography setup but lack the retopology, texturing and optimization expertise in-house. The handoff point that tends to work best is right after Step 1 (capture) a studio with existing lighting discipline can usually produce consistent, well-overlapped source photos on its own, while the reconstruction, cleanup, UV work and GLB optimization stages benefit from specialized software experience that is harder to build up for occasional use.