Aeion Spatial + Luma AI
Luma AI is the leader in photorealistic 3D generation — using Neural Radiance Field (NeRF) technology + diffusion models. Aeion submits high-fidelity text-to-3D requests to Luma; outputs glTF/glb meshes ready for placement in any scene. Asset pipeline auto-processes (real decimation + LOD generation on the .glb output; texture compression isn't wired up yet). BYO API key supported. Included quota with the $29/mo flat plan; passthrough costs for heavy usage.
What Luma Provides
`
Setup
`
Generation Workflow
`
Prompt Engineering for Luma
`
Common Use Cases
`
FAQ
Luma: highest fidelity, slowest, most expensive. Meshy: faster, cheaper, slightly lower quality. Tripo3D: best for character/animation-ready topology. Pick the provider per request — there's no automatic routing today.
For most uses: yes. For top-tier feature film + AAA game: still need traditional CG. Aeion's AI generation is for rapid prototyping + mid-tier production + customization.
Luma's TOS allows commercial use of generated assets. Verify current terms at lumalabs.ai for your specific use case.
Each generation counts whether quality is good or bad. Refine prompts + try again; included quota typically covers iteration.
Luma improves continuously; current 2-5 min per asset is faster than 2023 (was 10-15 min). Expect further reduction.
Not supported through Aeion today — Aeion's Luma integration calls Luma's standard generation API, with no per-tenant custom/fine-tuned model configuration.
Limited. Luma's outputs are typically static meshes. For rigged characters, use Aeion's character creator + Tripo3D (better topology for rigging).
Yes. Upload 360° walkaround video; Luma reconstructs scene from multiple angles. Higher quality than single-image; more cost + time.
Luma returns glb (and USDZ or a Gaussian splat depending on input type); glb is what Aeion's asset pipeline processes end-to-end. Converting to FBX/OBJ afterward isn't a reliable path yet — see the glTF/USDZ integration page for current format-conversion caveats.
Yes. Luma processes prompts on their servers. For sensitive use cases, BYO with private API key; consider Enterprise on-prem AI models instead.