Production-Ready 3D. Powered by Your GPU.
Generate, inspect, optimise and package game-ready 3D assets using the GPU you already own. Keep your source assets local and own every stage of the production pipeline.
One predictable fee. Unlimited local runs.
Your assets are processed locally and remain under your control.
Your GPU, not a shared queue
3D Oasis pairs with the NVIDIA GPU already in your own machine and runs generation there. No shared cluster queue, no per-job metering of compute you already paid for.
- Generation runs on your paired Windows Compute Node.
- One predictable fee. Unlimited local runs.
- Optional cloud processing modes are planned for later — never required.
Sign in, pair, upload, generate
The browser is the studio. Your Windows PC is the processing engine. The two stay connected over a secure outbound connection — no inbound port on your network.
- Sign in and pair your Compute Node once.
- Upload a reference image from the browser.
- Generation happens on your GPU; progress streams back live.
Know what your hardware can handle before you commit
A short, safe benchmark profiles your GPU and recommends which generation models and settings are realistic for your hardware — before you spend a single generation run.
- Real hardware detection, not a guess from a spec sheet.
- Compatibility report with plain-language recommendations.
- Thermal and stability warnings before anything runs long.
From a reference photo to a game-ready asset
Upload one or more reference images and generate a base 3D model on your own GPU. Unseen geometry beyond what the image shows is inferred, not guaranteed reconstruction — you review and refine from there.
- Background removal and image preparation before generation.
- A real, versioned first pass — not a locked-in final result.
- Every generation stays on your machine unless you export it.
Inspect the way a technical artist actually works
Orbit, measure, and switch render modes in a viewer built for production review, not a toy preview panel.
- Solid, textured, unlit, normal, and wireframe render modes.
- PBR channel viewer, UV checker, and material inspector.
- Scene hierarchy, measurement tools, and turntable preview.
Leave with everything a pipeline needs, not just a mesh
A finished export is a complete archive: runtime asset, source textures, an editable scene, validation reports, and previews — packaged so it can be handed to another tool or another person.
- Manifests documenting exactly what's inside.
- Licence and provenance records travel with the bundle.
- Reproducible builds — rebuild runtime output without rerunning AI.
One canonical asset, three real export formats
Every export derives from the same internal representation, so the runtime GLB, the editable Blender scene, and the full-fidelity OpenUSD package stay consistent with each other.
- Runtime-optimised glTF/GLB for engines and viewers.
- An editable .blend scene with materials, lights, and cameras intact.
- OpenUSD for full-scene composition and interchange.
GPU-efficient textures by default
Runtime textures ship as Basis Universal-compressed KTX2, so assets load faster and use less GPU memory without you having to think about compression settings.
- Automatic channel packing to cut texture count.
- Every compressed texture is validated before export finishes.
- Raw, uncompressed source textures are still included in the bundle.
Your assets are processed locally and remain under your control
Source images, generated geometry, and textures stay on your Windows machine. The cloud layer only ever coordinates jobs — it never sees your files.
- No raw images, meshes, or textures leave your device by default.
- Training on your content is disabled.
- Any future cloud-processing option will be opt-in and clearly labelled.
Built for people who ship into an engine
Outputs are structured for real game-production workflows, not just pretty renders.
- LODs, collision meshes, and scale/pivot correction.
- Game-readiness reports before you import anywhere.
- Formats that drop into Unity, Unreal, Godot, or a DCC tool.
Why local-first, for now
| 3D Oasis (local) | Cloud generation | |
|---|---|---|
| Where processing happens | Your own GPU, on your own machine | A third-party server (not available in this version) |
| Asset privacy | Source files never leave your device by default | Requires uploading assets to a remote service |
| Cost model | One predictable fee, unlimited local runs | Typically metered per job or per minute of compute |
| Hardware ceiling | Limited by your own GPU's VRAM and speed | Scales with whatever hardware the provider offers |
Before you sign in
- Do I need my own GPU?
- Yes, for this version. 3D Oasis pairs with a Windows PC running a supported NVIDIA GPU and runs generation there. There is no shared cloud compute yet.
- Will this reconstruct anything hidden in my reference image?
- No. Geometry for parts of the object not visible in your reference image is inferred by the model, not guaranteed to match the real object. Review and refine before treating a generation as final.
- Do you train on my uploaded images or generated assets?
- No. Training on private user content is disabled by default, and your asset files are not uploaded to the cloud layer in the first place.
- What happens to my files if I stop paying?
- Your local files stay on your own machine regardless — that's the point of local-first storage. Account and billing details will be published alongside real subscription terms before commercial launch.
- Is this ready for AAA production pipelines today?
- Not yet, and we won't claim otherwise. Milestone-by-milestone, we're building toward a production-grade pipeline; current output quality reflects an early, actively-developed platform.
Pricing will be announced before commercial launch
3D Oasis is built around one predictable fee for unlimited local generation on your own GPU, with billing and entitlement rules to be published here once they exist. This section intentionally does not list numbers yet.
Own the Pipeline.
Sign in, pair your GPU, and see what your own hardware can produce.
Benchmark My GPU