How to convert a hand drawing into a 3D model using AI, step by step

What you need

• A drawing of one object. Paper, a drawing app or the Vi3W canvas all work.
• A browser (Chrome, Firefox or Edge). Nothing to install, and the generation runs on Vi3W's servers, not your graphics card.
• A free Vi3W account. Sign-up gives you 100 XP, and one model with the default settings costs 30 XP.

Everything below was done on November 9, 2026, with a drawing made in about a minute with a mouse. The screenshots are from that run.

Step 1: Draw the object, not the scene

Our drawing was deliberately rough: a box, a curved lid, a perspective side, two bands, two plank lines and a lock. Things that helped:

• A three-quarter view. One front face and one side tell the AI the object's depth.
• Closed outlines. A gap in the outline can be read as a hole.
• No shading, no labels. Arrows and notes get modelled as geometry.

If you are drawing on paper, use a dark pen or go over the pencil, and take the photo straight on in even light. A faint grey line on cream paper is the most common reason an upload "isn't recognised".

Step 2: Describe it in a few words

Type what the object is into the prompt bar. We used "A wooden treasure chest with iron bands and a lock". The words set the material and style. The drawing sets the shape. Without the words, a box with a curved top could be a bread bin.

Step 3: Generate the image and check it

Press Generate. Vi3W renders your drawing into a full-colour image with Nano Banana. Ours took 17 seconds. This is the step most people skip on other tools, and the one that saves the most time. Ask yourself:

• Is it the object you meant, from the angle you drew?
• Are the proportions right? Is the lid the right height?
• Is anything added that you didn't draw?

If the answer is no, redraw the part that was misread or change the words, and generate again. An image costs 10 XP. A wrong 3D model costs more.

Step 4: Generate the 3D model

Go to Image Gen, choose the 3D model and press Generate 3D. TRELLIS is the default and the cheapest. Vi3W removes the image background automatically before handing it over. Our chest arrived about 35 seconds later.

If the back or sides matter, add a side view (upload one or use Generate Views) before this step. A single front image means the AI invents everything it can't see.

Step 5: Look at the mesh, not just the texture

The 3D Gen tab has four views: Textured, Solid, Wireframe and Normal. Spin the model in each one.

• Solid shows shape problems a nice texture hides.
• Wireframe shows the topology. Raw AI output is dense triangles. Ours was 32,210 triangles and 23,149 vertices.
• Normal shows bumps and seams as colour.

Step 6: Export to where you work

Download the model as GLB for the web, Godot or three.js, FBX for Unity and Unreal, or OBJ for Blender and slicers. Vi3W retopologises the mesh before export, so it is lighter and easier to edit than the raw triangles above.

Next steps depend on the destination: Blender, Unity, Unreal Engine or 3D printing.

When the result is wrong

• It's the wrong object. Change the words before you change the drawing.
• It's flat or squashed. Draw a perspective view, or add a side view.
• Thin parts are missing. Thicken handles, straps and legs in the drawing. AI models lose anything thinner than a few pixels.
• The back is nonsense. Generate or upload a back view.
• It needs to move. Vi3W makes static meshes. Rig it afterwards (how).

Try it with your own drawing

Open Vi3W, draw one object or upload a photo of your sketchbook, and follow the six steps. The free XP covers about three complete models.