Image to 3D Model.

Local when you can. Mutual when you can’t.

Run the full TripoSplat model on this device when its WebGPU and storage qualify,
or send the same job to a free MutualGPU provider from this page.
Qualified devices can also opt in to help process another person’s TripoSplat job.

Try it out Learn how it works Why this matters

SYSTEM CHECK

Preparing your browser

Checking WebGPU and storage…

STEP 1 / IMAGE

Choose a starting image

YOUR INPUT

Use a transparent PNG or WebP, or prepare an opaque photo locally before generation. The processor removes the background, frames the subject, and creates the official 1024px TripoSplat input.

Or start from an image URLThe image host must allow browser CORS access.

No image selected yet.

OPTIONAL / LOCAL PROCESSING

Prepare a photo for TripoSplat

Opaque photos start background removal and the official subject crop automatically in this browser. Your pixels stay on this device; use the button to process again.

Choose an image to enable local processing.

Continue to browser engine

STEP 3 / RESULT

Your spatial scene

Waiting

A completed scene will appear here without leaving this page.

STEP 2 / EXECUTION

Choose where it runs

CHECKING

Use this device for private, browser-local WebGPU generation when it qualifies, or use MutualGPU without local GPU or storage requirements.

Execution mode Run on this deviceChecking WebGPU, model storage, and desktop requirements…LOCAL

Run on MutualGPUUse free community-contributed compute. No local WebGPU or 6.5 GB model cache required.NO LOCAL GPU REQUIRED

The result returns to this same viewer. Selecting remote mode sends the selected image to the MutualGPU exchange and assigned provider.

Engineering preview. A first local run may download about 6.5 GB and generation can take several minutes. MutualGPU wait time depends on network, service, and provider availability. Browser support and output quality vary.

Checking local model storage and MutualGPU configuration…

Local model server

Model CDN directoryUsed only for on-device runs and provider work. Must contain manifest.json and allow CORS.

MutualGPU connection

MutualGPU HTTPS APIDefaults to the public MutualGPU exchange. The requestor origin and result storage must allow this application.

STATUS

Checking browser compatibility…

ENGINEERING PIPELINE

What this browser is doing

Guided mode explains each stage in plain language. Technical details remain one click away.

01
Source imageInspect alpha and prepare local pixels.
READY
02
Model packageVerify manifest, integrity, and browser cache.
WAITING
03
ConditioningPreprocess → DINOv3 → Flux VAE.
WAITING
04
Flow sampler20 official steps · 40 CFG DiT calls.
WAITING
05
Spatial decodeOctree occupancy → Gaussian decoder.
WAITING
06
Scene handoffEncode PLY / .splat and swap the viewer.
WAITING

RUN REPORT

Copyable benchmark summary

Complete a generation to capture its timing and browser environment.

Privacy: local mode keeps image processing and inference on this device. MutualGPU mode uploads the selected image to the exchange and assigned provider. Provider mode receives untrusted requestor images and uses the local model cache. TripoSplat model artifacts come from the official Hugging Face package unless you choose another server.

OPTIONAL / SHARE YOUR GPUBecome a MutualGPU providerChecking

Qualified Chrome devices can receive one remote requestor’s TripoSplat assignment at a time. Accepted work runs on this browser and can download the same 6.5 GB model cache while sustaining heavy GPU and power use for several minutes. The requestor identity is not disclosed, but every activity line shows the assignment ID, phase, and progress. The provider key stays in this page’s memory and is never saved. Closing the page interrupts the session. Only enable this in an operator-controlled browser.

Provider keyIssued for one MutualGPU execution unit. Advertised machine profileChoose a profile this device can honestly sustain.

Checking local WebGPU, model, and storage requirements…

REMOTE PROVIDER ACTIVITY · LAST 120 EVENTS

Select any text, or copy the complete log. New activity never moves your reading position.

Provider is idle. Accepted assignments run on this browser for a MutualGPU requestor; each activity line identifies the remote task.

WHY THIS MATTERS

A monumental attempt to put CUDA-era 3D generation in reach of more people.

TripoSplat started as a CUDA-based 3D generation pipeline. With Codex GPT 5.6 Sol Ultra and GPT 5.6 Sol Max in Kiro, this project is porting that work to WebGPU—replacing the usual requirements for NVIDIA hardware, CUDA, terminal commands, environment setup, and local dependency installation with a browser-native workflow.

The goal is not a cosmetic port. It is an attempt to make advanced AI 3D generation radically more accessible: for GPUPoor MacBook users, students, independent developers, and anyone without an expensive GPU. When a browser passes the system check, they can generate a 3D asset locally on hardware they already own—no terminal, no installation, and no cloud GPU bill for inference.

This is still an engineering preview, with compatibility and quality under active qualification. But the direction is clear: keep powerful spatial-generation tools inspectable, local, and available to more builders.

Read the source, benchmarks, and open issues on GitHub