Skip to content

About StemStudio

What StemStudio is, who builds it, and how it stays free.

Last updated: 24 August 2026

What this site is

StemStudio is a collection of file tools that run entirely in the browser. Stem separation, file conversion, image and video compression, audio analysis, mastering, key and tempo changes — plus smaller things like character counts and text diffs. There are 64 of them at the moment.

All of them are free, and none of them ask you to register. There is no usage cap, no watermark, and no paywall on the export step. You open a page, pick a file, and finish the job there.

The tools are grouped by the kind of work they do rather than by file type: audio, files, images, text, documents, security, calculation, network and weather. The home page is the index. Drop a file on it and it shows you what can be done with that format first.

Why everything runs on your device

A typical online converter uploads your file to a server, processes it there, and sends the result back. That is easier to build and it makes heavy work fast — but from your side it means a copy of your file now sits on someone else’s computer. When it gets deleted, and who can read it in the meantime, is something you can only take on faith.

StemStudio does not have that structure at all. The separation models, the video compressor, the PDF merger — all of it runs inside the browser tab. Your file never leaves the page, so there is no deletion policy to believe. Nothing was handed over in the first place.

There is a cost. Speed depends on your machine, and the larger AI models will not load on a device with little memory. We do not hide that: every page states the expected processing time and the conditions under which a model is unavailable. A full list of what is and is not transmitted is on the "What is sent" page.

What it is built on

Stem separation uses Meta’s Demucs and Deezer’s Spleeter, converted to ONNX. Both are MIT-licensed pretrained models. They run through ONNX Runtime Web, using WebGPU where the browser supports it. The model files are downloaded to your browser straight from Hugging Face, not from our servers.

The rest of the tools are built out of what the browser already provides: Web Audio API for analysis and audio processing, Canvas for image conversion, WebCodecs for video compression, WebRTC for device-to-device transfers. Every dependency, with its copyright notice and licence, is listed on the Open Source Licences page.

How it stays up

Running costs are covered by advertising. Ads are how the tools stay free; there is no paid tier and no metered billing. Ad units are kept visually separate from the text and are not placed next to controls.

Some pages carry affiliate links to third-party products or services. Where they do, they are labelled as advertising. Going through one never raises the price you pay.

This domain used to distribute Tarasque, a Japanese input method for Windows. Distribution and sales ended in September 2026, and anyone who already bought it is being looked after individually. What this site publishes now is the browser tools alone.

About what is written here

Numbers on a page have to match the implementation. If a page says it targets −14 LUFS, the tool uses −14 LUFS. If a page says something can only be extracted with the 6-stem model, that page opens with the 6-stem model. We do not write copy that is more attractive than the thing it describes.

Limits are stated up front too. A bass boost, for example, has a ceiling past which the track distorts or the limiter takes the gain straight back off — pushed to the letter of the request, the request stops being met. So the ceiling comes first, and the setting goes as far as it usefully can inside it.

The blog exists to explain what actually happens when you use these tools. We do not rephrase the same article to cover more search terms.

Who runs it

One person, independently. Design, implementation, writing and illustration are all by the same individual.

At a glance

Operator
stemstudio
Address and phone
Disclosed without delay on request (by email, below)
Contact
contact@stemstudio.jp
Site
https://www.stemstudio.jp
Tools published
64 (audio, files, images, text, documents, security, calculation, network, weather, games)
Languages
Japanese and English

Related pages