# How to Turn a Character Image Into a 3D Model With AI

A character drawing, a selfie or a page from a character sheet can become a textured mesh in the browser. This guide covers preparing the picture, generating and inspecting the model, auto-rigging a humanoid, and exporting for Unity, Unreal, Godot, Blender or the web.

_Andreas Edesberg · 2026-11-09 · 3D Modeling_

---
You have the character already. It is a concept sketch in your sketchbook, a selfie on your phone, an illustration a commissioned artist sent back, or one panel from a character sheet you drew months ago. What you do not have is the mesh, and the traditional route to one runs through blocking, retopology, UV layout and texture painting before the character ever moves.

A single flat picture is now enough to get the first version of that mesh. This guide covers the whole path for a character specifically: what the picture needs to look like before you upload it, how to generate and inspect the model, how to get a skeleton into a humanoid so it can move, which export format to pick for each destination, and when a different Sloyd tool is the better answer.

## What "Character Image to 3D" Means

It means one 2D image of one character becomes a three-dimensional object you can orbit, light and export. [Sloyd Image to 3D](/image-to-3d) is an AI 3D model generator that runs in the browser: feed it a single photo, drawing or sketch and it returns a textured model that already carries its geometry, its UV maps and its textures. The sides the drawing never showed are reconstructed, which is the part that would otherwise take you an afternoon of modelling.

That is a different job from a parametric character builder. A builder hands you a human body and lets you move its proportions with sliders, so the identity comes from the parameters you set. An image-driven generation keeps the identity that is already in the picture: the silhouette, the costume, the hair, the hat brim, the pose. The rule of thumb for the rest of this guide is simple. If the likeness of a specific design matters, start from the image. If you need a clean, conforming human body, start from sliders.

Two more things worth getting straight before you upload anything:

- **It is one character per picture.** Use a picture with a plain, uncluttered background so the AI can separate the subject from its surroundings, because a clear single-subject picture works best. A drawing of two duelists around a campfire is three separate uploads: one cropped to each duelist, and one more for the campfire if you want the prop as well.
- **The output is a mesh, not a rig.** A textured model is what Image to 3D produces. Bones, skinning and motion come from the rigging and animation step further down this page, and skipping that step leaves you with a static statue.

## How to Prepare a Character Image

The reconstruction is only as good as what the picture lets it see, so this is where most of your result is decided. The converter accepts JPG, PNG and WEBP source images up to 20MB.

| Aim for | Avoid |
| ------- | ----- |
| The whole character in frame, head to feet | A crop that cuts the legs, hands or the top of the head |
| A plain, uncluttered background | A busy scene the character blends into |
| Even, flat lighting across the figure | Hard rim light, deep shadow or a blown-out highlight |
| A calm standing pose, arms slightly out from the torso | Arms folded, a hand in a pocket, a weapon held across the chest |
| Front or three-quarter view | A strict profile, or heavy foreshortening toward the lens |
| Matte fabric, leather, wood and painted metal | Chrome, wet surfaces, glass and transparent fabric |
| A sharp image with confident linework | Motion blur, soft focus or faint pencil |

Four of those rows are worth a sentence more.

**Occlusion is the expensive mistake.** Anything the picture hides has to be invented. A cloak thrown over one shoulder, a shield covering the chest or a hand tucked behind the back all reconstruct as a guess, so treat every hidden part as a risk rather than as something you can count on. Separating the limbs from the torso costs you one redraw and saves the cleanup. If the character does already exist as several views, [Multi-image to 3D](https://app.sloyd.ai/create-3d?mode=multi-view) takes multiple angles of one subject and improves coverage of the parts a single picture cannot show; it costs 20 credits per run on Sloyd Plus or Sloyd Pro.

**Flat light preserves shape.** Strong directional light bakes shadow into the texture and reads as geometry that is not there. An evenly lit figure gives the generation a clean read of where the volume actually changes.

**Matte beats reflective.** Shiny and transparent materials have no stable surface to reconstruct, so armour described and drawn as brushed steel lands better than mirror chrome, and a stained-glass cape lands worst of all.

**Pose for geometry, not for the poster.** The flying kick belongs in your splash art. Feed the model the standing version of the same character, then animate the kick later.

If you are coming at this from a sketchbook rather than a camera, the longer version of the drawing side of this workflow is in our guide on [turning concept art into a 3D model](/blog/from-concept-art-to-3d-model).

## The Step by Step in Sloyd

### 1. Upload the Character Image

Open [Image to 3D](https://app.sloyd.ai/create-3d?mode=image), drop in your JPG, PNG or WEBP, and select Generate. Nothing installs, because the whole flow runs in the browser.

### 2. Inspect the Mesh Before You Do Anything Else

Sloyd creates the geometry and textures automatically, and you inspect the generated model in the browser, reviewing its geometry, textures and proportions from every angle. Spend a moment on this rather than exporting straight away. Orbit to the back of the character, which is the side the drawing never showed, and look at the places where limbs met the body in the source picture.

Detail is set with the Polycount slider rather than in the picture. It runs from 3k to 500k triangles, with topology set to auto, quads or triangles. The top step, 500k (Ultra), is available on Sloyd Plus and Sloyd Pro, spends credits from your monthly allowance on every run, and locks topology to triangles and textures to 4K; free generations are capped at a lower polycount. If the character is heading to a modeller for retopology or edge-loop work, pick quads at one of the steps below Ultra. If it is going straight into an engine, leave it on auto.

When something reconstructs badly, the fix is almost always a better input rather than a different setting: redraw with the arm clear of the torso, flatten the lighting, or crop less tightly.

### 3. Auto-Rig the Humanoid

A character that stands still is a prop. For humanoid characters prepared in a T-pose or A-pose, Sloyd fits a skeleton automatically without manual joint placement or weight painting, which is the step that turns the mesh into something that can walk. This is why the pose advice above matters twice: an A-pose input gives you a better reconstruction and a better rig from the same upload.

The automatic rigging works best on humanoids, so a quadruped mount or a floating creature is worth inspecting closely before you build a set of motions on top of it. The mesh itself generates the same way either way. For background on what rigging is doing underneath, see [how to rig AI models for animation](/blog/how-to-rig-ai-models-for-animation-in-minutes).

### 4. Add Motion and Preview It

With a rig in place, the [Animate tool](/animations) applies motion such as walk cycles and idles, and you preview the result in the browser before you export. One rigged character can carry a set of motions, so walk, run, idle and attack live on the same model rather than on four exports.

Two things here cost credits instead of being unlimited. Text-to-animation creates motion from a prompt and requires a subscription. Video-to-animation maps a short full-body reference clip onto your character and is available on Plus and Pro. Both are paid out of your monthly credits, while the automatic rigging and the basic animation that sits on top of it are not.

### 5. Export for Your Destination

Open the export dialog, pick the Game Dev preset for an engine target, and download. The formats and the reasoning are in the next section. When you have applied multiple animations and want all of them in one export, choose Download Animations → All rather than the default current-animation-only export.

## When a Different Path Works Better

Image to 3D is the right tool when a specific character design exists as a picture. Three cases where it is not:

- **You need a conforming human body.** The [3D human model generator](/3d-human-model-generator) is parametric: 45 body sliders for height, shoulder width, waist, limb proportion and muscle definition, plus facial expression and pose controls, exporting as FBX or PLY with no account needed. Reach for it when you want a correct human you can dial in, not the likeness of a drawing.
- **You need a platform avatar shape.** Roblox avatars have their own tool at the [avatar creator](/avatar-creator). You pick a mannequin, describe the look, and get a fully textured Roblox avatar with AI texturing, exported as GLB for Roblox Studio. Every mannequin is skinned and the avatar shape stays unchanged, so the result stays compatible with the Roblox marketplace.
- **You want a quick personal character from a photo.** The [selfie to 3D character app](/apps/selfie-to-3d-character) renders a stylized 2D image from your photo. Send its output through Image to 3D before rigging or exporting.

## Export Formats That Matter for Characters

The Game Dev export preset offers GLB, FBX, OBJ, OBJ plus MTL, USD, USDZ and BLEND, and recommends GLB. Once the character is rigged and animated, the animation-capable formats are GLB, FBX, USD, USDZ and BLEND, so pick from that shorter list when motion has to travel with the mesh.

| Destination | Format to pick | Why |
| ----------- | -------------- | --- |
| Unity, Unreal or Godot | FBX, or GLB with a glTF importer | FBX imports with no setup; GLB is the smaller file |
| The web | GLB | Compact, single file, textures included |
| Blender for retopology or rig work | BLEND | Native file, quads preserved |
| A modeller that only reads classic formats | OBJ, or OBJ plus MTL for textures | Widely supported; plain OBJ carries geometry only |
| A USD pipeline | USD, or USDZ | Fits collaborative scene assembly |

For how a character export fits into the rest of an asset pipeline, the [game development use case](/use-case/game-development) page covers the full loop.

## Free Versus Paid

You can try the whole reconstruction step without paying. Without a subscription you get 1 free generation per day, with no account needed, which is enough to find out how your particular drawing behaves before committing to a plan. Free generations are capped at a lower polycount. Alongside that daily generation, the Guest tier lists downloading community models, unlimited in-browser template customization, GLB and OBJ as the standard export formats, and a personal-use licence.

What the Guest tier does not list is exporting a character you generated yourself. Sloyd Plus and Sloyd Pro are what unlock that, and using the result commercially: both plans include a commercial licence. On the paid plans, text and image to 3D and AI rigging and animation are unlimited and do not spend credits. The advanced steps do: Ultra generations of up to about 500K triangles with 4K textures cost 5 credits each, Multi-image to 3D costs 20 credits, and text and video to animation spend your monthly credit allowance. Compare the tiers on the [pricing page](/pricing).

## Start From the Picture You Already Have

The character you want is already drawn or already photographed. Prepare the picture so the whole figure is in frame on a plain background with the limbs clear of the torso, generate, inspect the back of the model in the browser, auto-rig it if it is humanoid, then export GLB or FBX for your engine.

[Open Image to 3D](https://app.sloyd.ai/create-3d?mode=image) with that picture, and if the result is a character you want to move, carry it straight into [Animate](https://app.sloyd.ai/animate).

## FAQ

### Can AI turn a character drawing into a 3D model?

Yes. Sloyd Image to 3D works in the browser: you upload one picture of the character, generate, and inspect the result from every angle. What comes back is a textured model with its geometry, UV maps and textures already built. One photo, drawing or sketch is enough, so you do not need a turnaround sheet, multiple camera angles or a scan. If you do have several views of the character, Multi-image to 3D is a separate, credit-based path on Sloyd Plus and Sloyd Pro that uses them all.

### What kind of character picture gives the best result?

A sharp image with the whole character in frame, even lighting and a plain background. A calm standing pose with the arms held slightly away from the torso reads far better than a mid-action pose, because every limb the drawing hides has to be guessed. Matte clothing beats reflective clothing, and bold linework beats faint pencil.

### Does the generated character arrive rigged and ready to animate?

No. Image to 3D gives you a static textured mesh. The rig is a separate step: for a humanoid character prepared in a T-pose or A-pose, Sloyd fits a skeleton automatically, with no manual joint placement or weight painting. After that the Animate tool applies motion that you preview in the browser before exporting.

### Which format should I export a character in for Unity or Unreal?

The Game Dev export preset offers GLB, FBX, OBJ, OBJ plus MTL, USD, USDZ and BLEND, and recommends GLB. GLB is the compact choice and FBX is the one that needs no importer setup. Rigged, animated characters export as GLB, FBX, USD, USDZ or BLEND.

### Can I turn a character image into a 3D model for free?

You get 1 free generation per day with no account needed, which is enough to see how your picture reconstructs. Free generations are capped at a lower polycount. Exporting your own generated character and using it commercially come with Sloyd Plus or Sloyd Pro, which also make text and image to 3D and AI rigging and animation unlimited.

### Should I use an image or a parametric character builder?

Use the image when the likeness matters, because the costume, the proportions and the face of the drawing are what you want to keep. Use a parametric human model generator when you need a clean, conforming human body you can dial in with sliders and you do not care about matching a specific picture.

---

Source: https://www.sloyd.ai/blog/how-to-turn-a-character-image-into-a-3d-model-with-ai