Text to 3D
Describe it, pick a concept, get the 3D model
Type what you want and HI3D.cc draws four concept images in about 10 seconds, free, up to 10 sets a day. Pick one and it becomes the input for the same image-to-3D pipeline as a photo: a shape preview in 2–3 minutes, then a 4K model with PBR textures in 4–12 minutes that looks like the concept you chose.
| Step | What you do | Time | Credits |
|---|---|---|---|
| Concepts | Describe one object; four images appear | ~10 s | 0 |
| Pick | Choose the design you want, or make four more | — | 0 |
| Shape preview | Rotate the clay model and check it | 2–3 min | 40 |
| 4K model | Starts by itself unless you stop it within 30 seconds | 4–12 min | 60 |
The 3D print and game-ready presets work the same way from a concept: 40 credits for a print mesh, 60 for a game-ready model, both without a preview step.
Why concepts first
We tested text straight to 3D. It changed the design.
In our first paid test we generated a shape preview and then the final model directly from the prompt below. They came back as two different chairs: the preview was a one-piece white chair with curved arms; the final had a walnut frame and separate mustard-yellow cushions. Measured surface to surface, they differed by 9.3% of the height on average.
From a single image, the preview and the final stay the same object: 0.64% difference for a full-body portrait, 2.35% for a mech robot. So HI3D.cc moves the choice to a picture you can see, and the model you approve is the model you pay for.
“A mid-century modern lounge armchair with a curved walnut wood frame, tapered legs and a mustard yellow fabric seat and back cushion, a single piece of furniture, product design”
Prompts
How to write a prompt that builds well
- One object. A 3D model is one thing, not a scene. Leave out rooms, landscapes and other characters.
- Name the materials and colours. “Weathered bronze”, “walnut and mustard fabric” give the textures something to follow.
- Describe the silhouette. Tall or squat, tapered legs, a round base, arms away from the body.
- Say “full body” and “standing” for characters, and hands apart from weapons if you plan to rig it.
- Skip lighting and backgrounds. We ask for a plain background and a centred object automatically.
| Weaker prompt | Stronger prompt | What changed |
|---|---|---|
| an owl | a weathered bronze owl statue on a round stone base | Material, style and a base to stand on |
| a knight fighting a dragon in a castle | a knight in plate armour holding a longsword, full body, standing | One subject, no scene, a clear pose |
| a cool chair | a mid-century lounge armchair, curved walnut frame, tapered legs, mustard cushions | Shape words and two materials |
| a robot | a chunky industrial robot with rounded armour plates, orange and grey, arms at its sides | Proportions, colours and a pose that rigs well |
Good fits
What text to 3D is good for
Good fits
- Props and set dressing for games and renders
- Furniture and product concepts to show a client
- Original characters, then game-ready and auto-rigged
- Figurine ideas to print
Better another way
- A real product you own: use a photo
- Exact dimensions: set them when you export for print
- Logos and lettering: use an extrusion tool
- Real people: needs their photo and consent
Questions
Text to 3D, answered
Is text to 3D free?
The concept images are: up to 10 sets of four a day, no credits. Building the one you pick costs the same as a photo: 100 credits for a full-colour 4K model, 40 for a print mesh, 60 for a game-ready model. New accounts get 40 credits, enough for a shape preview.
How long does text to 3D take?
About 10 seconds for the four concept images (7.9 seconds for two in our test), then the same times as a photo: a shape preview in 2 to 3 minutes and the 4K model in 4 to 12. Our text-based 4K run took 281 seconds.
Can I write the prompt in another language?
Yes. Prompts are translated into a clean English description of one object before the images are drawn, so Spanish, German, Chinese or Japanese all work.
Can I use my own concept art instead?
Yes: switch to the Image tab and upload it. Shaded renders and paintings of a single object work best; flat line drawings carry less information about depth.
Why do I have to pick an image before the 3D model is made?
Because a 3D model generated straight from text can change between the preview and the final. In our test the same prompt gave two different chairs. Picking an image first fixes the design, and a model built from one image keeps its shape from preview to final.
Can I describe a real person or a famous character?
Describe original designs. A model of a real person needs their consent, and characters and logos you don't own can't be used commercially even if the model is yours.
Have a photo instead? Image to 3D. Planning to sell what you make? Commercial use of AI 3D models.