Run AI models in your own browser
Not a demo of someone else’s API. The model itself is downloaded into your browser and executed on your hardware with WebGPU, so the prompt, the reply and the image never leave the machine. Load one, switch to flight mode, and keep going.
- Runs entirely in your browser
- Works offline
- No upload
- No sign-up
- No watermark
Loading the tool…
How it works
- Pick a model. Each one shows its download size and the hardware it wants before you commit.
- Wait for the download — once only; it is kept in your browser cache afterwards.
- Use it. Then, if you want the proof, turn off your connection and use it again.
Why nothing is uploaded
Every operation on this page is done by code running inside your browser tab, using the same engine that renders web pages. The file is read from disk into your tab’s memory, transformed there, and written back out as a download. It is never sent anywhere — not to us, not to a third party.
Verify it yourself
- Open your browser’s developer tools (F12) and select the Network tab.
- Load your file and run the tool.
- The only requests you will see fetch the tool’s own code — and, for a few heavy tools, their open-source engine from a public CDN — plus one small page-view ping to loreatec.jp (page address and title, nothing more). None of them carry your file.
Frequently asked questions
Is this really running locally, or is it calling an API?
Locally, and it is the easiest claim on this site to verify: load a model, switch to flight mode and keep using it. An API call cannot survive that. The only network traffic is the one-off weight download from Hugging Face, which you can watch in the Network tab.
Why are the downloads so large?
Because these are the actual models, not a thin client. Depth estimation is 30 MB, speech synthesis 175 MB, transcription 282 MB, image generation 1.9 GB, and the largest chat model 4.8 GB. Each is fetched once and cached; the size is shown next to every model before you start, and there is a button to delete them all afterwards.
What can the eight models actually do?
Two Gemma 4 models take text, images and audio and answer in text. Gemma 3 1B is a small text-only chat model for weaker machines. LCM Dreamshaper generates images from a prompt. Kokoro reads text aloud. Whisper transcribes audio. DETR finds and labels objects in a photo. Depth Anything estimates depth from a single image.
Do I need a powerful computer?
For the small ones, no: depth, detection and speech run on an ordinary laptop. Image generation and the large chat models want a real GPU and enough free memory — each model carries a tier badge so you can tell before downloading gigabytes. Without WebGPU everything still runs on the CPU, much more slowly.
Which licences do these models carry?
Six are Apache-2.0, one is MIT and Gemma 3 1B is under the Gemma Terms. All are open models used here for a free demonstration; each one is named on the page so you can read its card on Hugging Face.