Introducing MeshAura: TRELLIS.2 Image-to-3D, Pay-As-You-Go

MeshAura is live. Upload a single image — a photo, a product shot, a piece of concept art — and get a fully textured, production-ready 3D model back in minutes. No modeling, no rig of GPUs to babysit, no subscription. This post covers exactly what's available today and where we're taking it next.

Built on TRELLIS.2

Under the hood, MeshAura runs Microsoft's TRELLIS.2 — the current state of the art in open image-to-3D. It's a four-billion-parameter model that reconstructs geometry and full PBR materials (base color, roughness, metallic) from one image, with sharp features and clean topology. We handle the hard part — running it on cloud GPUs behind a simple web app and API — so you just bring an image.

If you've tried image-to-3D before and been underwhelmed, the input matters more than the tool: a single, clearly-lit object on a plain background is what makes the difference. We wrote a practical guide to it , and the app now includes an in-browser cropper and input tips to steer you toward good results.

What's live today

• Image to 3D in minutes. Upload, preview the result spinning in your browser, download it. A live progress timer shows you exactly what stage a generation is in.
• GLB, STL, and OBJ export. Every generation gives you all three: GLB for Unity, Unreal, Godot, Blender, and three.js; STL ready for a 3D-printing slicer; OBJ for legacy DCC pipelines.
• Full PBR materials.Models come textured, so they look right under your engine's lighting — not just in a preview.
• A developer API. Two endpoints — submit an image URL, poll for the result — so you can drop image-to-3D straight into a pipeline, a marketplace, or your own product. See the API docs.

Pricing that respects your wallet

MeshAura is prepaid and pay-as-you-go. You add funds once, and each successful generation deducts a flat rate — no subscription, no credit bundles to decode, no monthly bill quietly running while you're heads-down on something else. Generations that fail are refunded automatically, so you only ever pay for models you actually get. Not sure yet? The interactive demos on our homepage are real results you can spin around before you spend a cent.

What's coming: choose your speed and cost

Today every generation runs on a single high-end GPU tier tuned for quality. That's the right default, but different projects want different tradeoffs — a hobbyist iterating on 3D prints cares about cost, a studio running a batch cares about turnaround. So the next thing we're building is choice of GPU tier:

• A cheaper tier— the same model on more cost-efficient hardware, for when you're exploring and want the lowest price per model.
• A faster tier — the latest-generation accelerators for the shortest possible turnaround when you need results now.

These will run as independent serverless pools you pick from at generation time, so you dial in the speed-versus-cost balance that suits the job rather than being locked to one. Alongside that, we're working on higher-detail quality options and text-to-3D (via a text-to-image step in front of the pipeline). We'll announce each as it ships.

Try it

Bring an image of a single object — a prop, a vehicle, a character, a gadget — and see what comes out. It costs a couple of coffees to experiment, failed runs are free, and everything you make is yours to use commercially.