Meshy – AI 3D Model Generator
Design & 3DFreemiumBest for: Text-to-3D and image-to-3D asset creation for games, printing, and product prototyping
Browser-based AI 3D generator: text, image, or sketch to textured 3D assets with PBR materials, rigging, animation, and 3D-printing tools.
Founded 2023
What Is Meshy?
Meshy is a browser-based AI 3D model generator. You feed it a text description, a photo, or a rough sketch, and it returns a textured, animatable 3D asset. All generation happens on the vendor's servers, so a laptop with a decent connection is enough — no workstation GPU, no installed modeling software.
The core pipeline is two-stage: a model stage produces the base geometry, and a separate texture stage adds PBR maps. Around that sit AI texturing for uploaded meshes, auto-rigging with a large animation library, a 3D-printing toolset, an API, an MCP server for agent workflows, and a CLI. The flagship line is Meshy 7, launched August 2026 with an image-to-3D focus and the closest prompt-to-model fidelity so far; Meshy 6 and the faster 6 Lite fill out the range. Everything on this page reflects the vendor's official site, help center, and August 2026 changelog, checked on 26 September 2026.
How Generation Actually Works
You start with an input mode. Text-to-3D takes a written description and returns candidates in roughly a minute on typical tasks. The 3D Agent accepts chat instructions, an image, or even a sketch, and handles multi-step requests such as generating a model and then retexturing it. Image-to-3D reconstructs geometry from one or more reference images — for the 7 series this is the only input route, because those models dropped text input; text prompts on the 7 series route through the Agent instead.
Model choice matters more than the interface. Meshy 6 Lite, released August 2026 as the successor to the discontinued Meshy 5, is the free, fast option — the vendor's docs describe a draft preview in about eight seconds, aimed at rapid iteration on a watertight, clean-topology base. Meshy 6 (January 2026) raises geometry quality, with sharper hard-surface edges and more anatomically correct organic forms. The 7 series (7.1 current) delivers the highest detail and the closest reference fidelity, at the highest credit cost.
- Model stage: 25 credits on the 7 series, 20 on Meshy 6, 10 on 6 Lite
- Texture stage: 10 credits at 2K/4K resolution, 15 at 8K
- In the 3D Agent, image-to-3D with texture costs 30 credits (15 with Smart Topology)
- Failed tasks refund credits; costs are identical across all plans
Who It Helps Most
The value differs a lot by audience, and it pays to be realistic about who gets the most out of it.
Game developers get fast asset drafts and a genuine path to engine preview: generate, texture, rig, animate, export GLB or FBX, then let artists clean up the hero pieces. Indie devs in particular can block out whole levels with 6 Lite while spending real attention on the few models the camera lingers on. 3D artists use it as an ideation engine and as a shortcut for texturing and rigging assets they modeled elsewhere — the AI texturing tool accepts FBX, OBJ, STL, GLTF, or GLB uploads up to 100MB. 3D-printing users get the deepest toolset: watertight output, a printability check with automatic repair, auto split, multi-color support, and slicer plugins from Bambu Studio to Ultimaker Cura. Product designers can turn a photo or concept sketch into a reference prototype for visualization or a first physical print, then iterate cheaply before commissioning anything. XR/VR prototypers and concept artists benefit most from image-to-3D: a good reference sheet becomes a blockable mesh in minutes instead of hours of manual blockout.
Game-Art Workflow: From Prompt to Engine-Ready Asset
Treat it as the front half of a pipeline, not the whole pipeline. A practical path for a playable game asset looks like this:
Concept → reference. Gather one strong reference image, or a multi-view sheet from the built-in AI image generator, which produces views optimized for reconstruction. → Generate. Use the 7 series for hero assets and 6 Lite for blockouts. → Clean topology. Export with Quad Mesh for cleaner edge flow, then decimate in Blender. Guideline budgets from the vendor's help center: 100–500 triangles for a mobile background prop, 1,000–3,000 for a mobile hero, 500–5,000 for a PC environment prop, 10,000–50,000 for a console hero. → Texture. Keep the PBR pack — albedo, normal, roughness, and metallic maps drop straight into Unity's URP or Unreal's material system. → Rig and animate. Auto-rigging covers humanoid and quadruped skeletons, and paid plans include 600+ motion presets. → Export. GLB for Unity and web, FBX for Unreal and DCC pipelines; OBJ as a universal fallback. → Polish in-engine. Build LODs yourself — exports are a single mesh with no LOD variants — and recalculate normals in Blender if geometry imports dark.
Every step after export is documented by the vendor's own help center, and that manual part is the difference between a nice render and a usable game asset. Official plugins for Blender, Unity, Unreal Engine, Roblox Studio, and 3ds Max move models between apps without a file round-trip.
- Quad Mesh export is the right choice when you plan to retopologize, rig, or deform a model
- Use the polygon slider in the download panel to reduce density before exporting
- No LOD auto-generation exists — set up LOD Group in Unity or static-mesh LODs in Unreal yourself
- Merge by Distance and Fill Holes in Blender handle most surface holes and floating geometry
Practical Prompts to Start With
The five prompts below follow the workflows documented in the vendor's help center and community patterns — not a test session on our side — so treat them as starting points and budget a few credits for iteration.
Game prop: "Rusted steampunk telegraph machine, weathered brass, dark iron panels, desktop-sized hard-surface prop for a narrative game, no glass." Expect to run a Smart Topology pass and a decimate before real use; hard-surface props respond well because small parts stay distinct. Stylized character: upload a clean character concept sheet and use image-to-3D on the 7 series with "bulky goblin guard, round helmet, big pauldrons, cartoon proportions." Reference alignment is where the 7 series earns its credits, but the character still needs retopology before deformation-heavy animation. Product prototype: "Cordless electric screwdriver, matte two-tone shell, glowing power ring, clean product-render look." Generate, retexture an uploaded CAD or photo at 2K, and print-test one unit before approving the direction. 3D-printable object: "Modular plant-pot base with snap-fit saucer, 8 cm diameter, watertight, no floating parts." Run the printability check, auto-repair, then export STL for your Bambu Lab and verify snap tolerance in the slicer rather than trusting the first print. Environment blockout: "Low-poly desert rock arch cluster, four formations, silhouette-first for a level blockout." Use 6 Lite for ten-credit drafts, keep the strongest arrangement, then redo hero pieces on the 7 series.
Keeping Characters Consistent Across Shots
There is no character-lock feature here — input is prompts and images only — so consistency has to be engineered into the workflow. It works well for one clear reason: image-to-3D hugs the reference image.
Build a character sheet first. Render or draw one consistent front view of the character, then use the multi-view AI image generator to expand it into several angle images optimized for reconstruction. Feed those views into multi-view image-to-3D on a paid plan (the free tier omits multi-view). Keep the same model version for every asset in a project — the vendor's docs note that outputs vary between versions, so mixing 6-series files with 7-series files is how characters drift. For repeated appearances, generate the hero once and reuse that mesh with different poses instead of regenerating: animation presets plus custom poses cover most scene needs, and LOD variants from the same base keep the character consistent and cheap. Expect variations rather than clones — the same prompt and reference produce similar but not identical models — so lock the version you like and stop iterating on it.
Common AI Game-Art Failures and Their Fixes
AI-generated 3D output is usually a starting asset, not a finished production asset, and knowing the typical failure modes saves the most time.
Broken or missing topology shows up as holes, non-manifold edges, or floating islands, with thin parts — blades, straps, fabric edges — as the usual culprits. Fix it by regenerating with a simpler, more focused prompt, or in Blender with Merge by Distance and Fill Holes. Inconsistent proportions and asymmetry are most visible in faces and hands; a better reference sheet helps more than any prompt tweak. Texture artifacts such as seams, smeared details, or muddy normal maps usually improve at 4K or 8K, or by running a fresh texture pass instead of a full regenerate. Backside geometry and interior volumes inflate triangle counts with zero visible payoff; Smart Topology or a decimate pass handles that. Non-game-ready topology — dense uniform polygons, triangulated meshes that deform badly — is why Quad Mesh export is recommended for anything that will deform, followed by retopology in Blender or Maya. Overly dense output for props and environments should be cut with the polygon slider or Decimate before import. Overlapping or stretched UVs break lightmaps and texture bakes, so re-unwrap anything that needs clean UVs for production. And when models import dark, it is usually just inverted normals — Select All, then Mesh → Normals → Recalculate Outside fixes it in seconds.
Concept Art to Production Asset: Where the Tool's Part Ends
The honest concept-to-production story is a relay race, and the baton passes at export. The tool reliably shortens the first half — concept, reference, generation, then a first pass at topology, texture, rig, and animation. The second half stays manual. Deforming characters need retopology tuned to their skeleton. LOD chains and texture atlases are built by hand. UVs get unwrapped for lightmaps and disc loading. Rig weights get painted where deformation matters. Engine import — collisions, interactions, materials — is yours. None of that is a criticism; it is the difference between generating a model and shipping a production asset, and any tool that claims to skip it entirely is overselling. The platform's value is the front half: turning days of blockout work into minutes so artists can spend their hours where they still matter.
For 3D-printing users the baton passes earlier. The print suite — watertight generation, printability check, automatic repair, auto split, multi-color AI optimize, and STL/OBJ/3MF export — covers most mesh-side work. What stays with you: slicer settings, support strategy, wall thickness, and the physical test prints that prove a design actually works on your printer.
Credits, Pricing, and Licensing in Practice
Credits are the currency here, and every task spends a fixed amount that depends on the model, task type, and resolution — there is no universal "one model costs X credits" rule. Plan size matters because it bounds how much iteration you can afford. The Free plan grants 100 credits monthly (reset on the 1st, no carryover) with no card required, enough to learn the loop. Its real limits: only 6 Lite models can be downloaded, capped at ten per month, and free outputs carry a CC BY 4.0 license, so commercial use requires crediting the vendor.
Individual paid plans start at Pro ($20/mo, 1,000 credits, unlimited downloads, private ownership, API access, multi-view), scale through Premium ($40/mo, 3,000 credits, faster queue, more retries) to Ultra ($100/mo, 8,000 credits, highest individual priority). Higher tiers are the better cost-per-credit, which matters if you generate daily. Studio ($70/seat/mo) is the team plan, starting with a 5,500-credit pool and growing in 1x/3x/5x multiples; Enterprise is quoted custom with SSO and dedicated support. New Pro/Premium/Ultra subscribers get 50% off their first month, and annual billing carries its own discount. Purchased credit packs never expire while subscription credits reset monthly. On licensing, paid plans grant private asset ownership with no attribution, and the free tier keeps a CC BY 4.0 allowance instead. Refunds are generally not offered, per the help center, so size conservatively.
Figures change, so verify before committing: the official pricing page, the help-center plan comparison, and the credit-cost breakdown are the authoritative sources as of September 2026.
Pricing at a Glance
Compare Plans & Pricing
Free (100 credits/mo); Pro $20/mo (1,000 credits); Premium $40/mo (3,000); Ultra $100/mo (8,000); Studio $70/seat/mo; Enterprise custom
Free
$0
Try the full workflow with 100 monthly credits, no card required.
- 100 credits every month
- 10 free downloads for Meshy 6 Lite models
- CC BY 4.0 license with attribution
- One-task queue, 20 animation presets
- No API or multi-view on the free tier
Most Popular
Pro
$20/mo
The most common pick for solo creators and indie devs.
- 1,000 credits per month
- Unlimited downloads and private ownership
- API access and 600+ animation presets
- Multi-view image-to-3D and community model downloads
- 10 concurrent tasks, higher queue priority
Premium
$40/mo
High-volume individual workflow with a faster queue.
- 3,000 credits per month
- 30 concurrent tasks and 12 free retries
- Everything in Pro, plus early access to new features
- 4 free Creative Lab designs per day
Ultra
$100/mo
Maximum credits for demanding individual workloads.
- 8,000 credits per month
- 100 concurrent tasks and 40 free retries
- Highest individual queue priority
- 12 free Creative Lab designs per day
Studio
$70/seat/mo
Team workspace with a shared credit pool.
- Monthly credit pool from $70 for 5,500 credits
- 60-task queue and 24 free retries
- Team management with multi-seat support
- Private asset ownership and limited API access
Enterprise
Custom
Custom contracts with SSO and dedicated support.
- Customizable credit volumes and throughput
- SAML SSO and API asset retention
- Highest API request rate
- Dedicated support and security review
Meshy vs Tripo vs Spline
Direct generator rivals are closer than the marketing suggests, so think in terms of workflow fit rather than winners.
Tripo is the most direct competitor: text-to-3D and image-to-3D, 8K PBR texturing, part-level segmentation, rigging, and engine plugins, with a free tier (200 credits/mo) that stays public and non-commercial. Where this platform leans into printing — watertight validation, repair, auto split, slicer plugins — Tripo leans into raw speed and the 8K texture ceiling. It also ships the MCP server and a broader plugin line-up across engines and slicers, while Tripo counters with part-based editing that is not available here. Neither one has a trivially clean topology story for organic characters, so benchmark identical prompts in both before choosing.
Spline is not really a competitor — it is a different job. Spline is a collaborative 3D scene builder for interactive web experiences: you compose, style, and ship scenes with AI assists, and its native habitat is the browser rather than a game engine or a slicer. If your goal is an interactive product page or a WebGL showcase, Spline is the better fit; if your goal is meshes you own and export, this tool is on-topic and Spline is not. A common pairing: generate hero assets with image-to-3D, then bring them into Spline for the interactive presentation layer.
Tripo
Design & 3DFreemium
Best for: Game-ready 3D asset generation at speed
AI 3D model generator that creates game-ready assets from text or images in seconds, with 8K PBR texturing, auto-rigging, and plugins for major engines.
Review: Strengths and Limitations
The browser-first approach is the headline: a usable, textured prop goes from prompt to downloaded file without leaving the tab, and the plugin ecosystem means results land where you already work. The printing suite is genuinely unusual — this is the only mainstream generator in the review group that ships repair, split, and multi-color tooling as part of the core product. Credit economics are fair for iteration-heavy workflows, and failed tasks refund. On the other side, the free tier's download restrictions and CC BY 4.0 attribution are real constraints, the 7 series' image-only input surprises people who expect text-to-3D, LOD work stays manual, organic topology still needs an artist's pass, and refunds are effectively off the table. Size the plan for your iteration rate, test on the free tier first, and treat the tool as the fast front half of a longer pipeline.
Pros
- Text-to-3D, image-to-3D, texturing, rigging, and animation all run in the browser
- Print-specific tooling: watertight meshes, printability check, auto repair, and auto split
- Official plugins for Blender, Unity, Unreal Engine, Roblox Studio, 3ds Max, and the main slicers
- API, MCP server, and CLI cover scripted and agent-driven pipelines
- Free plan is genuinely usable: 100 monthly credits and 10 Meshy 6 Lite downloads
Cons
- Free-plan downloads are limited to Meshy 6 Lite models (10 per month); Meshy 6 and 7 files need a paid plan
- Meshy 7 and 7.1 are image-to-3D only — text prompts must go through the 3D Agent
- Free-tier outputs are CC BY 4.0, so commercial use requires crediting the platform
- No automatic LOD generation; deforming characters still need manual retopology
- Monthly credits reset and subscriptions are effectively non-refundable
Best For
Recommended use cases and scenarios where Meshy shines.
Frequently Asked Questions
Common questions about Meshy, answered.
What is Meshy?
A browser-based AI 3D model generator that creates textured, animatable 3D assets from text prompts, images, or sketches, with PBR texturing, auto-rigging, animation, game-engine exports, and a 3D-printing toolset.
Is Meshy free?
Yes — 100 monthly credits, no credit card. It covers the full workflow, though downloads are limited to ten Meshy 6 Lite models per month and free outputs carry a CC BY 4.0 attribution license.
Can I sell models that I make with Meshy?
Yes. Paid plans grant private ownership, so you can sell or distribute your models freely with no attribution. Free-plan outputs use a CC BY 4.0 license, which permits commercial use as long as you credit Meshy.
Why can't I download a model I generated?
On the Free plan, only models generated with Meshy 6 Lite are downloadable, and they count against a 10-download monthly limit. Meshy 6 and 7 models show a paywall until you upgrade to a paid plan.
Which file formats are supported?
Downloads are available as FBX, OBJ, USDZ, GLB, STL, and BLEND. For AI texturing, uploads accept FBX, OBJ, STL, GLTF, and GLB up to 100MB per file.
How do credits work?
Every task spends credits, and the cost depends on the model, task type, and resolution — there is no universal one-model cost. Subscription credits reset monthly, purchased credit packs never expire, and failed tasks refund their credits.
Is Meshy good for 3D printing?
Yes — printing is one of its strongest areas. Output is watertight, and the built-in tooling covers printability checks, automatic repair, auto split, and multi-color optimization, with STL, OBJ, and 3MF export plus plugins for Bambu Studio, Creality Print, OrcaSlicer, and Ultimaker Cura.
How does Meshy compare to Tripo or Spline?
Tripo is the closest direct rival: similar generation, 8K textures, and part-level editing, but without this product's printing suite or MCP server. Spline is a different tool — an interactive browser-based 3D scene builder for web presentations rather than owning and exporting meshes.
Reviews & Ratings
Loading reviews...
Explore More Design & 3D Tools
Browse the full AI Tools Vault directory to compare Meshy with every design & 3d tool, or line up your shortlist side by side.
Similar Tools
More Design & 3D tools you might like
Canva AI
Design & 3DFreemium
Best for: Quick social & presentation design
Magic Studio AI design suite for creating graphics, presentations, and social posts in minutes.
Figma AI
Design & 3DFreemium
Best for: AI-assisted UI/UX design & prototyping
Figma AI generator turns text prompts into editable UI designs, wireframes, and prototypes inside Figma — plus AI search, smart editing, and design-to-code tools.
