Buying guide · Generators
Best Text-to-3D AI Tools
Which text-to-3D AI tool to use when you start from a prompt: documented text routes, credit costs, model limits and licenses for Meshy, Tripo, Rodin and more.
If you start from a written prompt, Meshy is the best-documented hosted choice: its web app has a Text to 3D flow with a published price of 20 credits on Meshy 6 (10 on Meshy 6 Lite). The catch is that model names change quickly: Meshy’s API changelog added Meshy 7.1 for Text to 3D on September 18, 2026, and the web-app price list we read names Meshy 6 and 6 Lite for text and says Meshy 7.1 is not available for Text to 3D there, so check which model your text form uses before you budget. This page picks by text route; for image input see best image-to-3D tools.
How the picks were made
We compared documented text routes, per-run credit prices, free limits, and licenses. The page makes no claim about how often a prompt gives a usable model on the first try; we have not measured it and found no source that publishes prompts and files behind such a number. Picks are judgments based on how completely each route’s price, limits and rights are published. See the methodology and, for writing prompts that help any tool, text-to-3D prompts.
| Tool | Price | Text route | Cost of a text run | Best for |
|---|---|---|---|---|
| Meshy | Free; Pro $20 | Text to 3D in the web app and API (preview, then refine) | Web 20 credits (Meshy 6), 10 (6 Lite); API preview 20 | First paid text-to-3D plan with published costs |
| Tripo | Not confirmed | Text to 3D, Text to Image (API) | 10 to 20 credits (API, $0.01 each; H3.1 no or standard texture, detailed and 8K are 30 and 40, P2 is 100 to 130) | Scripted pipelines; consumer plans unconfirmed |
| Rodin / Hyper3D | Creator $30 | Text-to-3D listed; basic exploration on Free | API 0.5 credit base | Detail-oriented generations |
| Sloyd | Plus $15 | Text to 3D (unlimited on paid plans per pricing page), LLM-guided creation | Subscription; LLM-guided creation 10 credits | Parametric props from a sentence |
| 3D AI Studio | Not confirmed | Text and image to 3D across several models | 20 to 60 credits by model (FAQ; its pricing page says 20, its homepage 25 to 35) | Trying models on one prompt |
| Hunyuan3D 2.1 | Free, own GPU | Text-to-3D tag on the model page | None locally | Self-hosted |
| TRELLIS (v1) | Free, own GPU, at least 16 GB | Text models (342M to 2.0B parameters) | None locally | Self-hosted text input |
Meshy: the documented default
Meshy publishes its text price, a refine step, and a policy of not charging for failed generations. Its pricing page lists Free at 100 credits a month, Pro at $20 for 1,000. The free tier limits downloads to untextured Meshy 6 Lite models, so a free text prompt will not give you a textured download. Meshy’s API changelog says meshy-5 was retired on October 10, 2026 (requests naming it now return an error, though existing meshy-5 models stay available), that meshy-6-lite was added for Text to 3D on September 11 (5 credits for the mesh via API, 15 with texture), and that meshy-7.1 was added for Text to 3D on September 18 and is what the latest alias resolves to. The web-app and API prices differ (web 10 credits for 6 Lite, API 5; Meshy has not reconciled them), so confirm in your account which model and price your text prompt uses. One model-lifecycle note belongs in any text-to-3D budget: models retire, so results from an old prompt may not reproduce. See AI 3D model releases.
Weaknesses: Free-tier download restrictions, paid-plan API output deleted after three days on non-Enterprise plans, and a documented conflict between its docs and Terms on credit expiry. Details in the Meshy review.
Tripo and Rodin: documented, partly unconfirmed
Tripo’s developer documentation lists H3.1 Text to 3D at 10 to 20 credits without texture or with standard texture (detailed and 8K are 30 and 40; P2 is 100 to 130) and Text to Image at 5 to 20 at base quality, which suits a text-then-image-then-3D workflow. Its consumer plan pages were blocked to us, so we state no plan price. Rodin lists text-to-3D among its features and, per its pricing page, lets Free users generate before confirming spend, but gives no free-credit amount. See Meshy vs Tripo.
Sloyd: text into parametric templates
Sloyd lists text-to-3D, and a credit price for LLM-guided creation (10 credits). Its homepage describes “AI-guided parametric templates”; we infer that this favors predictable props over unusual creatures, but that is our reading of the design, not documented output quality. Free Guest use is one text-to-3D generation per day and personal use only. See the Sloyd review.
Open models: a text route needs an image or old weights
TRELLIS.2, the current Microsoft model, is image-to-3D; the original TRELLIS has text models, and its README recommends generating text-to-3D by first making an image. Hunyuan3D 2.1’s Hugging Face tags list both image-to-3D and text-to-3D. Hardware and license details are in open-source AI 3D generators and free AI 3D generators.
What each pick means for the destination
A text route returns whatever the model’s export gives you, so the destination decides the pick as much as the prompt does. Meshy’s docs list GLB (textures embedded) and FBX (separate textures) among eight formats; Godot’s manual recommends glTF 2.0 and Unity’s recommends FBX, so Meshy covers both. Tripo’s text route is documented only for the API: its API formats are documented, Studio’s are unconfirmed. For printing, a text-to-3D result still has to be checked in a slicer; format and printability are separate questions, covered in printing guidance. If you plan to sell the result, free-tier outputs carry restrictions: Meshy Free is CC BY 4.0 with credit, and Sloyd Guest is personal use only.
Practical advice
A two-step route, writing the prompt as an image first and then running image-to-3D, is the workaround the original TRELLIS README suggests for text-to-3D. Whether it beats a direct text route is unmeasured; do a small trial on your own subject matter. The broader shortlist is best AI 3D model generators.
What evidence would change the pick
A published per-prompt result set with files, covering several tools on the same prompts, would replace the documentation-based reasoning above. We have no such set yet. Back to best-of guides.