Generate 3D objects from text prompts in seconds.
3D & AR
Luma Genie is the text-to-3D generator built by Luma AI, the San Francisco lab that spent 2022-2023 turning neural radiance fields into a mainstream tool with the Luma AI iPhone capture app, then followed it in 2024 with the Dream Machine video model. Genie sits inside that lineage as the "type a sentence, get a 3D mesh" product — a specialist for the moment in a concept-art or game-jam workflow where you need a rough hero prop in the next ten minutes and you would rather not open Blender to model it from scratch.
The product started life in early 2023 as a Discord bot that shipped free during beta and got surprisingly usable within a year. In 2024 it graduated into the main Luma AI web platform at lumalabs.ai/genie, kept its low-friction feel, and picked up a proper mesh-quality dial, better texture output, and Genie 2.0's much sharper geometry. In 2026 the product is a stable, mid-tier offering — not the frontier of generative 3D, not trying to be, and better for it. Meshy has pulled ahead on the pipeline features that game studios actually pay for (PBR maps, retopology, rigging). Genie has stayed closer to its concept-art roots and its parent company's real business, which is video.
The one-line positioning: Luma Genie is the text-to-3D pick when you want a fast, free-feeling on-ramp to generative 3D — not a production asset pipeline. It is the tool a 2D designer opens for the first time and enjoys, the tool a game-jam team spins up for placeholder props, the tool a solo indie hits for a mood-board scene when the internal budget is zero. If you are shipping a AAA game or building a Unity asset store business, you have outgrown Genie by lunch on the second day. If you are exploring what generative 3D even feels like, this is the least intimidating door in the category.
For AIQORA readers evaluating the current 3D generation wave — Luma Genie, Meshy, Spline AI, Kaedim, Rodin — Genie is the "start here" answer. You will know within a weekend whether generative 3D is a real fit for your workflow, and you will know it without paying anything.
Genie's feature surface is deliberately narrow. Luma's engineering weight in 2026 sits on Dream Machine, and Genie inherits the discipline of a product built by a team that ships a lot.
Text-to-3D generation. Type a prompt, get four mesh candidates in roughly 30-60 seconds each on the current model. The typical output is a single hero object — "a stone gargoyle wearing a top hat," "a low-poly diner in the desert," "a coffee mug with dragons carved into the side" — around 20k-100k triangles depending on subject complexity, with a baked color texture. Sharpness on organic shapes has improved substantially over the 2023 Discord-era output; hard-surface geometry is still where the model works best.
Image-to-3D generation. Upload a reference image and Genie reconstructs a 3D object approximating the subject. Cleanest on centered objects with clear silhouettes — a shoe, a chair, a helmet. Worse on messy scenes, transparent materials, or reflective surfaces. Reasonable for turning concept sketches into rough turntables.
Multiple mesh candidates per prompt. Every generation returns four variants at once. You pick the one closest to your intent and either export it or use it as a seed for a re-roll. The "four options every time" default is a small UX choice that matters — it turns prompting from a hit-or-miss lottery into a directable process.
Downloadable formats. GLB, OBJ, USDZ, and STL export from the web dashboard, which covers the workflows most users bring: real-time engines (GLB/OBJ), Apple ARKit (USDZ), and 3D printing (STL). No FBX in 2026, which will annoy Unity users who have not switched to GLB pipelines.
Textured meshes with baked color. Textures come baked into the mesh — vertex color or a single diffuse texture depending on model output. There is no separate PBR channel breakdown (roughness, metallic, normal). If you want a proper PBR-mapped asset ready to drop into Unreal, use Meshy; if you want a textured prop for a shader you will tune yourself, Genie's output is enough.
Web dashboard with prompt history. Every generation persists in your dashboard so you can go back a month, re-download a mesh you dismissed, or re-roll a prompt with the current model. Small but real quality-of-life win — the Discord version had a much more scattered history.
Luma AI ecosystem integration. Your Genie account is your Dream Machine account is your Luma AI capture account. One login, one billing surface, shared credit pool on paid plans. If you are already using Dream Machine for video, Genie's inclusion in the same account is a genuine convenience.
Discord bot for casual generation. The Discord-first workflow is still supported for creators who prefer chat-based prompting or want to generate collaboratively in a server. Slower to iterate on than the web dashboard but well-loved by the original beta community.
What is not in Genie in 2026: no retopology (meshes ship as-is from the model), no automatic rigging, no PBR channel outputs, no scene composition (one object per prompt), no animation. All four gaps are intentional — Luma has decided Genie is a generator, not a pipeline.
Luma Genie shares billing with Dream Machine under the unified Luma AI plans. The pricing model is credit-based, and both products draw from the same monthly credit pool.
| Plan | Price (USD) | Best for |
|---|---|---|
| Free | $0 | Trying it out — modest monthly credits, watermark-free 3D exports |
| Standard | $9.99/month | Casual hobbyists — small credit pool shared with Dream Machine |
| Plus | $29.99/month | Regular users — meaningful monthly credit pool |
| Unlimited | $94.99/month | Heavy users — no per-generation cost during business hours |
| Enterprise | Custom | Studios needing bulk credits, commercial licensing clarity, API access |
The credit accounting matters. A single Genie generation costs a few credits (single digits), so a Plus subscription typically nets you a few hundred 3D generations a month — plenty for concept work but nothing like unlimited. A Dream Machine video generation eats an order of magnitude more credits, which means if you are actively using both products, the Standard tier will feel starved fast and Plus is the realistic starting point.
Compared to Meshy (which prices its Pro tier around $20/month for a dedicated 3D pool) and Spline (which prices Super at $9/month for a design tool that happens to have AI), Genie's pricing is fair for the standalone case and genuinely good if you are also generating video with Dream Machine. The Unlimited tier at $95/month is competitive with mid-range Meshy plans and delivers a much bigger monthly volume for teams doing repetitive 3D work.
Commercial licensing is included on all paid plans. The free tier permits personal use and small commercial use with attribution — the practical read is that a solo indie shipping a $2 mobile game can use free-tier output, and a studio should be on paid.
Pros
Cons
Concept artists and 2D designers testing 3D. A designer who has never opened Blender can generate a stone gargoyle in 45 seconds and drop it into a scene mockup. That workflow is where Genie earns its keep.
Game jam prototyping and placeholder assets. A 48-hour game jam team can stand up dozens of prop candidates in a morning and pick the ones that fit, saving the "real" modeling for after the jam ships. Placeholder assets that outlast the jam are a bonus.
Indie devs mood-boarding a scene. For solo devs building a small indie title, Genie is a mood-board machine — generate twenty prop candidates, pick five, model the finals yourself. Cheaper than commissioning concept art.
Educators and students learning 3D. Art students learning topology can generate an object with Genie, retopo it in Blender for practice, and end up with something they made themselves at both stages. Cheap way to learn the pipeline.
AR/USDZ social content. The USDZ export means Genie output drops into iMessage AR previews, Instagram AR, and Apple's Reality Composer with no conversion. For social AR creators, the pipeline is basically zero friction.
Studios wanting a low-cost brainstorm tool. Even studios that use expensive tooling for finals benefit from a cheap upstream ideation tool. Genie's free tier fills that gap without a procurement conversation.
Do not use Genie for: production game assets (Meshy), scene composition (Spline AI), retopo-required deformation work (any DCC), or high-fidelity photoreal props (Kaedim, Rodin).
Meshy — the game-dev-oriented alternative. Meshy leads in 2026 on PBR channel outputs, retopology, rigging, and animation features. If your endpoint is a Unity or Unreal build, Meshy is where you should be looking. Genie is a better mood-board tool; Meshy is a better asset factory.
Spline AI — different product entirely. Spline is a web-native 3D design tool with AI features layered on top. If you want to build an interactive 3D scene for a webpage, use Spline. If you want a single 3D object to drop into an existing pipeline, use Genie.
Kaedim — the "convert your 2D concept into a production-clean 3D mesh" specialist, priced for game studios. Human-in-the-loop pipeline delivers higher-quality retopology than any pure-AI product. Different budget tier.
Rodin (Deemos) — a rising Chinese-founded competitor with strong hard-surface and character-model output. Worth a look for creators comfortable with a less mature UI.
Alpha3D — for enterprise product visualization and e-commerce 3D catalogs. Different buyer, different pricing, different job.
Stability AI's Stable 3D / Stable Zero123 — open-source-adjacent generative 3D. Rougher outputs, more control if you self-host, useful for research.
For most AIQORA readers evaluating text-to-3D as a category in 2026: try Genie's free tier first, try Meshy's free tier second, and pick based on whether you value the "starts fast, feels light" experience (Genie) or the "actually pipes into my engine" feature depth (Meshy).
Sign up at lumalabs.ai/genie with Google, Apple, or email. If you already have a Luma AI account for the capture app or Dream Machine, the same login works — Genie shows up as a tab on the dashboard.
Start with three prompts in your first session. Try one hard-surface prompt ("a low-poly diner in the desert, isometric"), one organic prompt ("a stone gargoyle wearing a top hat, weathered granite"), and one abstract prompt ("a floating crystal shard, glowing blue core"). You will learn what kinds of subjects the model handles well and what kinds need coaxing.
For each generation you like, download the GLB and drop it into a viewer you know — Blender's GLB importer, Windows 3D Viewer, macOS Quick Look, or an online GLB viewer. Look at the topology from a wireframe view before you commit any workflow to Genie's output; you will see immediately whether a retopo pass is required for what you want to do.
A practical tip: prompt with the material language you want, not just the object. "Ceramic vase" and "hand-thrown ceramic vase with iron-oxide glaze" produce meaningfully different textures. The model responds well to material specificity and atmospheric context; it responds poorly to compound scene descriptions.
Is Luma Genie free? Yes — the free tier is real and includes watermark-free 3D exports for personal and small commercial use. Regular users graduate to Standard or Plus.
Can I use Genie output commercially? On paid plans, yes, with full commercial rights. Free tier permits small commercial use with attribution — read Luma's current terms if the project is meaningful revenue.
Does Genie output PBR textures? No — output is a single baked texture per mesh in 2026. If you need proper roughness/metallic/normal maps out of the box, use Meshy.
What formats does Genie export? GLB, OBJ, USDZ, and STL. No FBX support in 2026.
How does the mesh topology look? Generative — often ugly under a wireframe view. Fine for static scene props, bad for deformation. Plan a retopo pass in Blender if you intend to rig or animate the mesh.
Is there an API? Luma provides API access on Enterprise tiers for Dream Machine and, increasingly through 2026, for Genie as well. Solo devs and small teams should assume no API access on the paid tiers.
Is it Discord-only? No — the Discord bot is still supported but the primary interface is the web dashboard at lumalabs.ai/genie. Both share the same account and credit pool.
How does Genie compare to Meshy? Different products despite overlapping demos. Genie is a fast, light, on-ramp generator with a unified Luma AI account. Meshy is a deeper asset pipeline with PBR, retopo, and rigging aimed at game devs.
Luma Genie is the on-ramp to generative 3D in 2026 — fast, free-feeling, and honest about what it is. For a designer, artist, or indie dev discovering the category, it is the friendliest starting point available. For a game studio shipping a real product, it is a mood-board tool and nothing more, and that is not a criticism.
The Plus plan at $29.99/month is the realistic subscription tier for a serious hobbyist who is also using Dream Machine, and the credit pool is generous enough that most users will not feel starved. The Unlimited tier at $95/month makes sense for creators generating dozens of assets a week or teams sharing an account. The free tier is enough to answer the question "is this worth paying for?" without a credit card.
Where Genie stops being the sharpest answer: the moment your workflow needs PBR channels, retopo, rigging, or animation. That is where Meshy has spent 2025-2026 pulling ahead, and the honest recommendation is to migrate to Meshy — or to a hybrid where Genie handles brainstorm and Meshy handles finals. If your work is web-first and interactive rather than asset-first, Spline AI is a different product with a different job and probably the better fit.
Try the free tier. If you use it more than twice a week, the Plus subscription pays for itself the first time you spin up a scene prop mid-meeting instead of budgeting a modeling day. If you use it twice a month, stay on free. If you use it every day, you are probably ready to graduate to Meshy anyway.