Text to 3D
Turn a written description into a 3D model when you have an idea but no reference image.
Text to 3D works well for exploring concepts, prototyping, and trying several design directions. If you have a reference image, Image to 3D matches it more closely. To brainstorm in a chat before you generate, use 3D Agent. Meshy makes models for art and games, not CAD-precision engineering parts.
Generate a model from text
To generate a model from text, follow these steps:
-
In the left sidebar, click Model.
-
At the top of the panel, click the Text to 3D tab.
-
Select High Detail for a detailed mesh, or Smart Topology for fewer polygons and cleaner topology.
-
Choose the AI model or topology:
-
If you selected High Detail, in the model list, select Meshy 6 or Meshy 6 Lite.
-
If you selected Smart Topology, which uses Meshy T1, under Topology, select Quad or Triangle.
-
-
In Prompt, describe the object in up to 800 characters. A detailed prompt names these parts:
-
The subject: "a medieval castle tower"
-
The material: "stone walls with moss"
-
The art style: "low-poly game art style"
-
The scale: "roughly 3 meters tall"
-
-
Optional: Set Pose to A-Pose or T-Pose for a character you plan to rig. Pose needs a paid plan, and free accounts get a limited number of free tries. With Meshy 6, you can also turn on Split to split the model into parts, or set Number of Generations to get up to 10 models from one prompt. More than one generation needs a paid plan.
-
Click Generate. The panel shows the time and credit cost before you generate. A Meshy 6 model takes about a minute, and a Smart Topology model about 3 minutes.
-
In the viewer, rotate and zoom to check the model.
The result is a mesh without textures. From the toolbar, you can texture, remesh, or download it.
Settings and tradeoffs
The following table shows what each choice trades off.
| Setting | Options | Tradeoff |
|---|---|---|
| Model type | High Detail or Smart Topology | High Detail gives more detail and more faces. Smart Topology gives fewer faces and cleaner topology, which suits games. |
| AI model | Meshy 6 or Meshy 6 Lite | Meshy 6 gives higher quality and a stable structure for 20 credits. Meshy 6 Lite is faster and costs 10 credits. |
| Art style | Set in the prompt | "Realistic" produces high detail. "Stylized" or "cartoon" yields cleaner topology. |
Examples
A specific prompt gives the AI enough to work with:
-
Prompt: "a detailed steampunk pocket watch with gears visible through glass, brass and copper materials, realistic style"
-
Result: a clean mesh with accurate gear details and about 25,000 faces.
A one-word prompt leaves the AI to guess:
-
Prompt: "watch"
-
Result: a wristwatch with little detail.
-
Fix: name the type of watch, its materials, its style, and the features you want to see.
Troubleshooting
The following table lists common problems, their causes, and how to fix them.
| Problem | Cause | Fix |
|---|---|---|
| The model doesn't match the description | The prompt is too vague | Add details: materials, style, viewpoint, and colors. |
| The model has floating fragments | The prompt describes several separate objects | Describe one subject per prompt. |
| The mesh has too many triangles | AI-generated meshes are dense | Use Remesh to rebuild the topology. |
| Texture seams are visible | The UV layout has issues | Use AI Texturing to regenerate the textures. |
FAQ
- How does Text to 3D work in Meshy?
- Describe an object in words, and Meshy generates a 3D model from the description. You don't need a reference image.
- How do I write a good text-to-3D prompt?
- Combine the subject, its material, the art style, and technical details, for example "a medieval castle tower, stone walls with moss, low-poly game art."
- How many credits does Text to 3D cost?
- A Meshy 6 generation costs 20 credits, Meshy 6 Lite costs 10, and a Smart Topology generation with Meshy T1 costs 20. The panel shows the cost before you generate.
- Can I use Meshy 7.1 for Text to 3D?
- No. Meshy 7.1 generates from images, so it's available in Image to 3D and Batch Images to 3D.
- Can I use a negative prompt in Text to 3D?
- Yes. Add exclusions such as "no background elements" or "no text or labels" to remove unwanted features.
Next steps
-
To write prompts that produce better models, see Prompting best practices.
-
To add color and materials to your model, see AI Texturing.
-
To generate from a reference image instead, see Image to 3D.