3D Models

Create stunning 3D representations of your physical assets directly in the Tokentagged web app. Transform your real-world objects into digital twins with photogrammetry technology – as a classic textured 3D mesh, a photorealistic Gaussian splat, or both. You can also upload your own finished model.

How It Works

Our 3D model generation uses photogrammetry to reconstruct three-dimensional models from photographs. The quality of your final 3D model depends heavily on the quality of your photographs and the characteristics of your object.

Complete Tutorial

Create a 3D model in just 3 steps:

Step 1: Take Photos

Photograph your object from all directions. Move the camera around the object instead of rotating the object itself. Make sure the photos are very sharp—the sharpness of your images is crucial for a good result.

Number of Photos:

• Take approximately 40 to 200 photos
• Minimum required: 20 photos
• Maximum allowed: 200 photos
• For complex and large objects, take as many photos as possible from different angles
• Formats: JPG/JPEG and PNG (HEIC is not accepted – convert first); 12 megapixels are plenty, larger photos are downscaled to 12 MP on upload

Photo Quality is Critical

The individual images don't need to show the entire object, but they must overlap significantly. Each photo should overlap with adjacent photos by at least 60% so our software can stitch them together correctly.

Photography Tips

📸 Click to view detailed photography tips

The result depends on both the object and the quality of your photos. Please follow these essential tips:

1. Close-Up Photography

• Photograph from a close distance so the object largely fills the frame
• The entire object doesn't need to be visible in each photo
• Partial sections are fine—just ensure significant overlap between images
• This overlap is essential for our software to combine the images

2. Sharpness is Everything

• Extremely sharp photos are crucial for success
• Zoom into your photos to verify surface textures are clearly visible
• Use good lighting and correct camera settings
• Use a camera with high resolution and depth of field
• Blurry images will result in poor 3D models

3. Lighting Setup

• Use diffuse, even lighting
• Avoid creating shadows on the object
• Outdoors: Bright, overcast days work best
• Indoors: Use multiple soft light sources

4. Background Matters

• Don't use homogeneous backgrounds (like photo tents)
• Do use textured backgrounds like newspaper or environments with visual structure
• Background texture helps the software calculate positions accurately

5. Learn from Examples

• Browse high-quality 3D models in the Tokentagged gallery
• View the source images used for those models
• Study real-world examples to improve your technique
• Learn practical tips from experienced creators

6. Suitability Check

• Not all objects are suitable for 3D modeling
• Check if your object meets the criteria below before investing time

Suitable vs. Unsuitable Objects

✅ Suitable Objects	❌ Unsuitable Objects
Objects with texture and structure	Moving or animated objects
Rough or matte surfaces	Highly reflective objects (mirrors, polished metal)
Static, still objects	Transparent objects (glass, clear plastic)
Examples: Statues, clothing, trees, buildings, pottery	Uniform objects without texture or detail

360° View Alternative

Even if your object isn't suitable for a classic mesh (e.g., transparent or reflective items), you can choose the photorealistic Gaussian splat, restrict the display to the 360° rotation view, or upload a finished model. The 360° view works for almost all still objects and provides an excellent presentation alternative. See Mesh vs. Gaussian splat.

Step 2: Upload & Calculate

Once you have your photos ready, it's time to upload them and start the calculation process.

The Process:

1. In your project choose "New 3D model" → "Calculate from photos (photogrammetry)" and upload your 20–200 photos
2. The AI pre-check looks at your photos (downscaled copies are sent to the AI briefly and not stored there) and recommends the method – Mesh, Photorealistic (splat) or Both (same price, a few minutes longer) – and whether to limit the area around the object (recommended for single objects; off for rooms and whole scenes). You can accept the recommendation or choose yourself. If it warns that the result will very likely not be satisfying, you must confirm to start anyway.
3. Accept the terms and proceed to payment

Pricing: Because 3D model calculation requires significant computational resources on powerful servers, this service cannot be offered completely free. You'll be prompted to:

• Pay via PayPal, or
• Redeem a voucher code if you have one

After payment confirmation, the calculation automatically begins.

Processing Time:

• Usually between 10 minutes and 2 hours, depending on the number of photos (mesh and splat together take a few minutes longer)
• You'll receive a notification when processing is complete

Save Money

Calculating a model costs €4 (VAT included). If you're not satisfied with the result, you can restart the calculation with better photos for €2, or request a 50 % refund (€2) before finalizing – the model is then deleted. If the calculation fails with an error, it can be restarted free of charge up to two more times or €3.50 is refunded. No refund is possible if a voucher was redeemed.

Step 3: Edit & Finalize

In the final step, you can review the result and prepare your model for display.

Review Your Model:

• Inspect the 3D reconstruction from all angles
• Check for any missing areas or artifacts
• Evaluate overall quality

Crop & Trim:

• Choose a cuboid or cylinder as selection box to define which part of the model to keep (cropping is not yet available for Gaussian splats – use "Finish without edit")
• Move the selection box with cursor buttons to the area you want to preserve
• Everything outside the selection box will be trimmed away
• Ensure the "TOP" label on the selection box points upward for correct orientation

info

Model editing can be repeated and adjusted later if needed. You're not locked into your first edit.

Not Satisfied? Before finalizing, you have two options:

1. Restart Calculation: Upload better photos and recalculate at a reduced price
2. Request Refund: Get 50% of your payment refunded

Finalize: Once you're happy with the result, click "Finish" and your model will be processed and ready within a few minutes (preview image and trailer video are rendered automatically). If you do nothing, the model is finished automatically with a default crop.

Settings: choose which views are shown and whether the photos are published with the model ("Do not publish images"), the default view, and the display – both variants, mesh only or splat only – plus the start model (trailer and preview follow the start model).

Best Practices Summary

Essential Requirements

• Sharpness: This is the most critical factor—zoom in to verify texture details
• Photo Count: 40-100 photos for most objects; up to 200 for complex items
• Overlap: 60%+ overlap between consecutive photos
• Lighting: Diffuse, even lighting without harsh shadows

Camera Movement

• Do: Walk around the object with your camera
• Don't: Rotate the object while keeping the camera stationary
• Coverage: Capture top, middle, and bottom angles

Background Setup

• Do: Use textured backgrounds (newspaper, patterned surface)
• Don't: Use plain white or single-color backgrounds
• Reason: Background texture helps position calculation

Object Suitability

• ✅ Best results: Matte, textured, static objects
• ⚠️ Challenging: Shiny, reflective, or very small objects
• ❌ Difficult: Transparent, moving, or completely uniform objects

Technical Specifications

Supported Input:

• Image formats: JPG, PNG
• Minimum photos: 20
• Maximum photos: 200
• Recommended resolution: 12MP or higher

Output Formats:

• GLB (optimized for web viewing); Gaussian splats as GLB with the KHR_gaussian_splatting extension
• Interactive 3D viewer (HD on capable devices, otherwise SD automatically)
• 360° rotation view
• Optional: downloadable model files (see Downloads)

Processing:

• Cloud-based photogrammetry
• Advanced algorithms for texture reconstruction
• Automatic mesh optimization
• Quality validation

Uploading your own model

You don't have to calculate a model from photos: a finished mesh (glb, gltf, obj, fbx, stl, ply or a zip with textures) or a Gaussian splat (ply, spz, splat, glb) of up to 200 MB can be uploaded free of charge. See Own models & Gaussian splats.

Integration with Tokentags

3D models can be directly linked to your tokenized assets:

• Set as the primary animation_url in your NFT metadata
• Provide immersive viewing experience for buyers
• Increase perceived value and authenticity
• Enable virtual inspection before purchase

Learn more about linking 3D models to your assets in the Tokentag Setup Guide.

Common Issues & Solutions

Blurry or Low-Quality Result

• Cause: Photos weren't sharp enough
• Solution: Retake photos with better lighting and camera stability

Missing Parts in Model

• Cause: Insufficient photo coverage or overlap
• Solution: Add more photos from different angles, ensure 60%+ overlap

Holes or Artifacts

• Cause: Reflective surfaces or poor lighting
• Solution: Use diffuse lighting, matte spray for reflective objects

Model Too Large/Small

• Cause: Incorrect cropping
• Solution: Re-edit the model and adjust the selection box

Learning Resources

• Gallery Examples: Browse the Tokentagged Gallery for inspiration
• View Source Images: Click on any 3D model to see the photos used
• Community Tips: Join discussions with experienced creators
• Practice: Start with simple objects before tackling complex items

Pro Tip

Before investing in photographing a valuable item, practice with similar test objects. This helps you understand lighting, angles, and photo count requirements without risk.