Last updated: 2026-10-29
Triverse AI is a browser-based 3D generator that turns one image or a text prompt into a mesh, with HD Mesh output of up to 1.5 million faces and PBR texture maps up to 4K resolution, per the vendor. It exports six formats, including GLB, FBX and STL, and has a REST API that launched in December 2025.
About Triverse
Triverse AI is a 3D generator that runs in the browser and over a REST API, operated by Guangzhou Light Illusions Technologies Co., Ltd. (LightIllusions), the party its Terms of Service name as providing the service. The company's newsroom names Ping Tan, a professor of electronic and computer engineering at HKUST, as its founder, and the CraftsMan3D paper behind its HD Mesh engine lists authors from HKUST, LightIllusions and Adobe Research. The pages read for this record give no founding year or funding figure, so none is stated here.
The workflow starts from a PNG, JPG, JPEG or WebP image, or a text prompt, and splits into two mesh engines. Artist Mesh produces lightweight low-poly assets meant for game engines, while HD Mesh, available since March 2026, outputs watertight geometry for rendering and printing. A separate texturing step paints the four standard PBR maps onto an existing model, steered by a reference picture or a prompt. Credits meter every run, and the cost depends on the mesh type and whether textures are switched on.
A reference image can come from anywhere: a sketch, a photo, or concept art made in Midjourney, Flux or Ideogram. The vendor's own use cases point to game props, 3D-printed miniatures and XR assets, and the files import into Unity, Unreal Engine and Blender as standard formats; the pricing cards mention 3D platform plugins, but none of the pages read names a plugin for any of those three. Unity developers weighing AI help on the code side can read the guide to AI for coding Unity games.
Within the AI 3D generators category of the media generation group, Meshy and Tripo count as the closest hosted rivals, Rodin adds agent and add-on integrations, and Hunyuan 3D covers running a model on your own hardware. Spline suits interactive web scenes and Marble suits explorable environments, so both solve a different job than a single-asset generator.
Pricing
Free is $0 with 300 credits a month. Pro is $9.99, Max $19.99 and Ultra $49.99 a month on monthly billing; the annual toggle shows $7.99, $15.99 and $39.99 a month billed yearly at $95.9, $191.9 and $479.9. Pro, Max and Ultra advertise a 50% first-month price that renews at list.
Extra credit packs for subscribers run from $8 for 500 credits to $40 for 5,000. On the vendor's own model counts, one Standard model costs about $0.075 and one Ultra Mesh PBR model about $0.175 at the Pro list price, and a Standard model costs $0.12 to $0.24 from the packs. Read from triverse.ai/pricing in both billing views on 29 October 2026; the vendor's llms.txt still shows an older table (a $39.9 Max plan, 100 free credits) and was not used.
| Tier | Monthly price | What it includes |
|---|---|---|
| Free | Free | 300 credits monthly recharge, about 20 Standard or 8 Ultra Mesh PBR image-to-3D models by the vendor count, 1 task in the queue at low priority, No free retries per model, Assets are public under CC BY 4.0, Asset downloads and storage listed as unlimited during the free trial |
| Pro | $9.99/mo | 2,000 credits monthly recharge, about 133 Standard or 57 Ultra Mesh PBR models, 10 tasks in the queue with high priority, 3 free retries per model, Batch to 3D generation, Assets are private and customer-owned |
| Max | $19.99/mo | 5,000 credits monthly recharge, about 333 Standard or 142 Ultra Mesh PBR models, 20 tasks in the queue with higher priority, Same retries, batch generation and ownership terms as Pro |
| Ultra | $49.99/mo | 20,000 credits monthly recharge, about 1,333 Standard or 571 Ultra Mesh PBR models, 50 tasks in the queue with the highest priority, Priority customer support, on top of what Pro already includes |
Key Features
- Image to 3D: Takes one PNG, JPG, JPEG or WebP image and returns a 3D model; per the vendor's pricing FAQ, a Standard model without textures costs 15 credits and an Ultra Mesh PBR model costs 35.
- Text to 3D: Builds a model from a typed prompt, with an optional prompt assistant that expands a short idea with 3D terms and lighting setups before generation.
- HD Mesh: Offers a face-count choice of 50K, 500K or 1.5M for watertight, manifold geometry aimed at rendering and 3D printing; the vendor says it has been available since March 2026.
- Artist Mesh: Produces lightweight, quad-friendly low-poly meshes for real-time use; the vendor quotes about 20 seconds per asset against about 3 to 5 minutes for HD Mesh.
- AI texturing: Adds Albedo, Roughness, Metallic and Normal maps to a mesh from a reference image or prompt, at resolutions up to 4K, without manual UV unwrapping.
- Export formats: Exports GLB, OBJ, STL, 3MF, FBX and USDZ, and the homepage FAQ says exports cost nothing extra.
- REST API: Offers a REST API with Bearer-token keys, WebSocket progress updates and a pricing-quote endpoint (added 2 June 2026) that returns a task's credit cost before it runs.
- Earnable credits: Lets free users earn credits: 150 per invited friend who completes a generation (capped at 1,500 a month), plus one-time grants of 200 for Discord and 100 each for X and YouTube.
Pros
- Paid plans keep every asset private and customer-owned, and the pricing cards say so in plain words rather than leaving it to the legal terms.
- Each plan lists how many Standard and Ultra Mesh PBR models its credits cover, so a budget can be checked before subscribing.
- The mesh method behind HD Mesh is described in a public paper (arXiv 2405.14979) with HKUST, LightIllusions and Adobe Research authors, so the approach can be read, not just advertised.
- API changes are logged with version numbers and release dates in public docs, so an integration can be audited against the changelog.
Cons
- Free-plan assets are public under CC BY 4.0, so private or client work needs Pro or higher.
- Subscription credits reset each billing cycle and do not roll over; only purchased packs persist, and those are sold to subscribers only.
- The Terms of Service (30 January 2026) say output is for reference and must not be used commercially unless explicitly permitted, while the homepage FAQ says commercial rights are included; get the permission in writing before shipping client work.
- The privacy policy sets no retention period and says de-identified uploads and outputs may be used to improve the service unless you object through the contact route the policy lists.
- The operator is registered in Guangzhou, and outside the United States the Terms are governed by PRC law, which some procurement teams will want reviewed.
- No independent ratings were found: Trustpilot has no page for the domain and search surfaced no G2 or Product Hunt listing, so the testimonials on its homepage are the vendor's own.
Data Handling
- Training-data policy
- The privacy policy says uploaded Content and the matching output may be used to improve the service after encryption and strict de-identification, tells users who do not consent to contact the vendor, and lists a right to object to personal data being used to train its models.
Frequently Asked Questions
What are Triverse's pricing plans in 2026?
Triverse has four plans: Free, Pro at $9.99 a month, Max at $19.99 and Ultra at $49.99, carrying 2,000, 5,000 and 20,000 credits. Yearly billing lowers the three paid rates by roughly a fifth, according to the pricing toggle. The first month of each paid plan is discounted 50%, then renews at list, so read the renewal line at checkout. Unused subscription credits expire each cycle.
Does Triverse have a free plan?
Yes. The Free plan needs no credit card and lists 300 credits a month, which the vendor counts as roughly 20 Standard runs or 8 Ultra Mesh PBR runs. It allows one queued task at low priority, gives no free retries, and makes every asset public under CC BY 4.0. The pricing card says monthly recharge while its FAQ text describes daily top-ups, so check the cadence on your own account.
What are Triverse's closest competitors?
[Meshy](/hub/tools/meshy) adds rigging and animation and ships mobile apps, which the Triverse pages read here do not mention. Choose [Tripo](/hub/tools/tripo) for quad Smart Mesh, auto-rigging and Blender or Unity plugins. [Rodin](/hub/tools/rodin) fits teams that want an MCP server for agents, and [Hunyuan 3D](/hub/tools/hunyuan-3d) suits anyone who wants to run the model on their own hardware.
How does Triverse compare to Meshy in 2026?
On price, Triverse's top plan costs less than Meshy's $70 Studio, and its entry paid plan sits within a cent of Meshy's $10 Starter. Meshy covers rigging, animation, iOS and Android, while the Triverse pages read here show only web and API access and split output into a low-poly Artist Mesh and a watertight HD Mesh. Choose Meshy for rigged characters or mobile work, and Triverse when game props or printable figures from a single image are the whole job.
What does it take to start using Triverse?
Sign up at triverse.ai (the Free plan asks for no credit card), open Studio and pick Artist Mesh for a low-poly asset or HD Mesh for a dense one. Upload a PNG, JPG, JPEG or WebP, or type a prompt, then download the result as GLB, OBJ, FBX, STL, 3MF or USDZ. For automation, create an API key in the API dashboard and send it as a Bearer token to api.triverse.ai/api/v1.
Top Alternatives
- Meshy: Pick Meshy when rigging, animation or a mobile app matter; pick Triverse when you want separate low-poly and watertight HD mesh modes at a lower top-plan price.
- Tripo: Pick Tripo when part segmentation and rigging in one workspace decide it; pick Triverse when you want plain file export and a Free plan whose public output carries a CC BY licence.
- Rodin: Pick Rodin when an MCP server for agents and Blender, Unity or Unreal add-ons decide the purchase; pick Triverse for a cheaper entry plan and a file-based workflow.
- Hunyuan 3D: Pick Hunyuan 3D if you want to run the model yourself; pick Triverse if you would rather rent a hosted studio and leave the GPU to the vendor.