Currently in Beta

Dumb AI

Image to 3D. On your machine.

Turn any image into a production-ready 3D model in seconds — completely offline, fully secure, and with no usage limits. Powered by Microsoft's open-source TRELLIS 2.0.

See How It Works
.GLB
.OBJ
.PLY
How It Works

From image to 3D in three steps

01

Upload an Image

Drop any single RGB image — a product photo, concept art, sketch, or reference.

02

AI Generates 3D

TRELLIS 2.0 encodes your image into a structured 3D latent space, generating geometry and textures simultaneously.

03

Export & Use

Download your 3D model in GLB, OBJ, or Gaussian Splatting format. Import directly into your pipeline.

Why Dumb AI

Built for privacy, speed, and scale

Fully Local & Secure

Runs entirely on your machine. No data leaves your network — ever. Your assets stay yours.

Unlimited Generation

No API limits, no per-model fees, no usage caps. Generate as many 3D models as you need.

Fast Pipeline

Image to production-ready 3D in ~10-15 seconds on a high-end GPU. Batch processing supported.

Multiple Export Formats

Export to GLB, OBJ, or 3D Gaussian Splatting (.ply). Ready for game engines, web, and AR/VR.

Powered by Microsoft TRELLIS 2.0

TRELLIS 2.0 is the latest evolution of Microsoft's open-source 3D generation model. Building on the original's rectified flow transformers and sparse 3D convolutions, v2.0 delivers sharper textures, better geometry, and faster inference — setting a new benchmark for image-to-3D quality.

Released under the MIT license, TRELLIS 2.0 gives you full commercial freedom. We've wrapped it into a polished, easy-to-use desktop experience — no terminal commands, no Python setup, no dependency wrangling.

Hardware Compatibility

What you need to run it

TRELLIS 2.0 requires an NVIDIA GPU with CUDA support. AMD and Apple Silicon are not currently compatible.

GPU

NVIDIA with 16 GB+ VRAM (CUDA required)

RAM

32 GB minimum

OS

Linux (native) / Windows (WSL2)

Storage

~20 GB for models + dependencies

GPU Compatibility

BestNVIDIA RTX 5090
32 GB VRAMFastest generation, maximum resolution
BestNVIDIA RTX 5080
16 GB VRAMFull speed, all features
GreatNVIDIA RTX 4090
24 GB VRAMFull speed, highest resolution
GreatNVIDIA RTX 4080 SUPER
16 GB VRAMFull speed, all features
GoodNVIDIA RTX 4070 Ti SUPER
16 GB VRAMSmooth with default settings
MinimumNVIDIA RTX 4060 Ti 16GB
16 GB VRAMFunctional with optimizations

GPUs with less than 16 GB VRAM (e.g. RTX 3060 12 GB) are not recommended. CUDA 11.8+ and Python 3.10 required.

Gallery

See it in action

Coming soon

Coming soon

Coming soon

Coming soon

Coming soon

Coming soon

Want to try Dumb AI?

We're onboarding early users for our beta. Book a call and we'll walk you through setup on your machine.

info@realityrift.co