Step-by-step game asset guide

How to Turn Pixel Art Into a Voxel Game Asset

Convert transparent pixel art into an editable voxel volume, reduce unnecessary colours and export one optimised surface mesh.

Updated 27 August 2026 · Reviewed against the current FORM 3D workflow

Typical time: 20–35 minutes
Reading time: 7 minute read
Difficulty: Beginner
Result: A compact coloured voxel model without one cube object per pixel

What this guide covers

A naive pixel-art conversion creates one cube for every visible pixel. That is easy to understand but wasteful: internal faces remain hidden, adjacent faces cannot merge and the scene may contain thousands of objects. FORM 3D instead stores a compact voxel grid and builds one exposed-surface mesh.

The source image still matters. Clean transparency, a deliberate palette and a bounded resolution produce better topology and a more coherent voxel style than feeding a large antialiased illustration into the converter.

Before you begin

Transparent PNG pixel art or a small hard-edged sprite
A target maximum grid between 32 and 64 pixels
A decision about whether depth should be flat, layered or reconstructed from three views

Prepare the pixel-art source

Use nearest-neighbour scaling and avoid blurred edges. Semi-transparent antialiasing creates ambiguous cells, so either clean the alpha edge or set a transparency threshold that removes those pixels.

Crop unused transparent space.
Use a small intentional colour palette.
Separate moving limbs or accessories into separate source images.
Check the sprite at 100% zoom before conversion.

Convert the image into voxels

Paste or upload the image and choose Pixel art → 3D.
Set Maximum grid to 32 for simple icons, 48 for moderate detail or 64 only when the silhouette needs it.
Use Exact colours for carefully authored pixel art.
Use Reduce palette when compression or gradients created many near-duplicate colours.
Use Single material for an untextured silhouette that will be coloured later.
Adjust the transparency threshold until background cells disappear without cutting the sprite.

Edit depth and layers

Select the voxel object and open its layer editor.
Use Add and Remove to repair the current depth slice.
Use Paint to change palette assignments without changing shape.
Use Box fill for large rectangular areas.
Duplicate depth layers to build thickness.
Mirror X, Y or Z when the design is symmetrical.

Reconstruct from front, side and top silhouettes

Three-view intersection keeps only cells that are visible in all supplied silhouette constraints. It is useful for blocky objects, but it cannot infer hidden concavity, texture or anatomy. Align and scale all three references before starting the worker.

Add visible references assigned to front, side and top views.
Align each image to the same ground level and approximate scale.
Use high-contrast silhouettes with matching proportions.
Run Front + side + top → voxels.
Inspect every depth layer and repair missing or excessive cells.

Why greedy meshing matters

Greedy meshing removes internal faces and merges adjacent coplanar faces with the same colour. A solid block can therefore export as six quads instead of thousands of cube faces. The visual voxel grid remains, but the game engine receives a much smaller mesh and node hierarchy.

Keep Merge surfaces enabled for the normal export path.
Use flat shading to preserve crisp voxel lighting.
Split parts only when they need independent animation or materials.
Review the exported triangle count rather than assuming every voxel is expensive.

Export the voxel asset

Choose a game engine profile in Export.
Review palette size, triangle count and object count.
Run Safe optimisation before Strong optimisation.
Verify and export GLB.
Check filtering and material settings in the target engine so the result remains visually crisp.

Frequently asked questions

Does every pixel become a separate cube?

No. Pixels become cells in one editable volume. Export meshing removes hidden faces and merges compatible exposed faces into one compact surface mesh.

What pixel-art resolution should I use for a voxel model?

Start at 32 pixels on the longest side. Increase to 48 or 64 only when the silhouette loses essential information. Resolution multiplies potential cells across width, height and depth.

Can one image create accurate voxel depth?

One image can create an extrusion or a manually edited volume, but it cannot reveal hidden depth. Front, side and top silhouettes provide better constraints, followed by manual layer cleanup.

Apply this workflow in the editor

Your project remains editable and stored locally on this device.

Open the studio