Image to 3D

Turn concept art, a photo, or a sketch into a 3D model that matches your visual reference.

An image gives the AI a clear target, so the result matches your design more closely than a Text to 3D prompt. If you don't have an image yet, generate one with AI image generation, or brainstorm in a chat with 3D Agent.

---

Generate a model from an image

To generate a model from an image, follow these steps:

1. In the left sidebar, click Model.

2. At the top of the panel, click the Image to 3D tab.

3. Select High Detail for maximum detail, or Smart Topology for an optimized, game-ready mesh.

4. In the model list, select a model. For High Detail, Meshy 7.1 - Flagship gives the most detail and precision.

5. Under Image, click the upload area, or drag and drop or paste an image. Use a PNG, JPG, JPEG, or WEBP file up to 20 MB. A clear subject on a simple background, at 512 × 512 pixels or larger, gives the best result.

6. Optional: Set the generation options. The panel shows only the options the selected model supports:

   • Multi-view (Meshy 7.1): add images of the same object from other angles. Multi-view needs a paid plan, and free accounts get a limited number of free tries. For more information, see Add views with Multi-view.

   • Resolution (Meshy 7.1): choose how much geometric detail Meshy generates. For more information, see Meshy 7.1 and resolution.

   • Poly Count (Meshy T2): set the target polygon count.

   • Split (Meshy 7.1 and Meshy 6): split the model into parts as part of the generation. You can't turn on Split and Texture together. For more information, see Auto Split.

   • Texture (not with Meshy 6 Lite): texture the model as part of the generation.

   • Pose: generate a humanoid character in an A-Pose, a T-Pose, or a Custom pose. Pose needs a paid plan, and free accounts get a limited number of free tries.

   • Image Enhancement (Meshy 7.1 and Meshy 6): let Meshy redraw your image for a cleaner, more complete model. It's on by default and helps most with uneven lighting or a busy background. Details can change, and generation takes longer.

7. Click Generate. The panel shows the time and credit cost before you generate. A Meshy 7.1 model takes about a minute.

8. In the viewer, rotate the model and check it from all angles.

When the model looks right, you can texture, remesh, or download it from the toolbar.

---

Smart Topology

Smart Topology generates meshes for game development, real-time rendering, and web 3D. It offers two models, as the following table shows.

What you get:

• Native part segmentation: the model comes split into clean, accurate parts. You can select, retexture, or animate each part on its own instead of editing one fused mesh.

• Fast iteration: Meshy T2 generates a mesh in about 2 seconds, so you can try several versions in one sitting.

• Poly-count control: with Meshy T2, set Poly Count anywhere from 100 to 15,000 polygons, from light web and mobile assets to more detailed real-time models.

Why it matters:

• Cleaner topology deforms more predictably, so rigging and animation behave better.

• A low, controlled poly count keeps assets light enough to use in your engine without a performance hit.

• Separate parts make later edits easier, such as swapping a material or adjusting one component.

Choose High Detail instead when you want maximum surface detail for renders or showcases and topology isn't a constraint.

---

Add views with Multi-view

With Meshy 7.1, turn on Multi-view to add images of the same object from other angles. Multi-view needs a paid plan, and free accounts get a limited number of free tries. Meshy combines all the views into one more complete model. Next to your main image, you can add up to three more views. If you don't have them, Meshy can generate them from your main image.

• Angles to use: front, side, back, and three-quarter views.

• Consistency: show the same object in every image, under similar lighting.

• Benefit: the back and sides come out more accurate, because the AI guesses less.

---

Set a custom pose

Custom Pose sets the exact stance your character is generated in, instead of the pose in your input image.

To set a custom pose, follow these steps:

1. Upload a character image.

2. Turn on Pose, and then select Custom. The Customize Pose dialog opens.

3. Start from one of the 6 preset poses, or drag and rotate each joint to adjust the pose.

4. Click Send to Custom Pose. If you close the dialog instead, Meshy discards your changes.

5. Click Generate. Meshy creates the character in the pose you set.

Use Custom Pose when:

• Your project needs a specific pose, such as an action stance or a sitting pose, and your reference image shows a different one.

• You want the same character in several poses without uploading a different image for each.

• You need a T-pose or A-pose for rigging in Animate, but your concept art shows a dynamic pose.

Tips:

• For the best results with Animate, which auto-rigs the model and applies animation presets, set the pose to T-Pose or A-Pose.

• Custom Pose applies to humanoid characters. Other subjects keep the pose shown in the input image.

---

Settings and tradeoffs

The following table shows what each setting trades off.

---

Meshy 7.1 and resolution

Meshy 7.1 is the Image to 3D model built for close alignment between your input image and the generated model. It handles:

• Organic characters with expression, anatomy, and skin folds.

• Multi-part objects, where each part keeps its position and neighboring parts stay separate.

• Engraved lettering and relief clean enough for CNC and laser engraving.

Meshy measures alignment on three things: overall proportion, spatial distribution, and surface details. Alignment applies to both single-image and multi-view input.

Resolution sets how much geometric detail Meshy 7.1 generates. It affects geometry only, not textures. The Ultra options generate the geometry at higher precision from the start, rather than adding detail to a finished model. The following table compares the options.

Ultra 2K and Ultra 4K are available on paid plans. On paid plans other than Enterprise, Ultra 2K is selected by default, so a Meshy 7.1 generation costs 25 credits unless you select Standard. If you generate for 3D printing, review the model before you print, because fine detail can print differently from a Standard model.

---

Examples

A clean reference gives an accurate model:

• Input: a clear front-view photo of a sneaker on a white background, with Meshy 7.1 and High Detail.

• Result: an accurate 3D sneaker with correct proportions, a detailed sole pattern, and clean textures.

A cluttered scene confuses the AI:

• Input: a photo of a full scene with several objects and a cluttered background.

• Result: the AI merges the objects into one distorted mesh.

• Fix: crop the image to a single object, remove the background, or use a cleaner reference.

---

Troubleshooting

The following table lists common problems, their causes, and how to fix them.

---

FAQ

How do I convert a photo into a 3D model?

In the left sidebar, click Model, and then upload a clear image in Image to 3D. Select Meshy 7.1, set any options, and click Generate. A model takes about a minute.

What image works best for Image to 3D?

A clear single subject on a simple background, at 512 × 512 pixels or larger. For a busy background, keep Image Enhancement on.

Is Image to 3D more accurate than Text to 3D?

Yes. A visual reference gives the AI a clear target, so Image to 3D matches your design more closely.

Can I generate a model in a specific pose?

Yes. Use Custom Pose to start from one of 6 presets or adjust each joint. Choose a T-pose or A-pose if you plan to rig the model.

Can I use multiple photos of the same object?

Yes. With Meshy 7.1, turn on Multi-view and add up to three more views, such as side, back, and three-quarter views, for a more complete model.

---

Next steps

• To generate without a reference image, see Text to 3D.

• To explore ideas in a chat before you convert concepts to 3D, see 3D Agent.

• To generate several models from different images at once, see Batch Images to 3D.