About TranscriptBalance

Updated 10 October 2026

TranscriptBalance is a free web app that turns audio and video into text right in your browser. It is built by Jeremy Heindrichs, a student in Belgium, alongside his studies. This page explains what it does, why it costs nothing and which principles it follows.

What is TranscriptBalance?

TranscriptBalance converts audio and video files to text and transcribes lectures, seminars and meetings live. The Whisper speech recognition model runs in the browser on your own device: your audio is not uploaded, and you need no account. The app is free, has no limit on minutes or files, and its interface is available in German, English, French and Dutch.

What TranscriptBalance can do

Transcribe files: several audio and video files at once, in 99 languages, detected automatically or chosen by you, with four Whisper models from “Flash” to “Precise”.
Tell speakers apart: who speaks when, with names you choose. More on this: Transcribe interviews.
Export: as TXT, Word (DOCX), Markdown, JSON or CSV, and as SRT or VTT subtitles, with or without timestamps and speakers.
Transcribe live: the audio of a browser tab, the screen or the microphone, with an AI chat, a live summary and an overview at the end. More on this: Live lecture transcription.
Summarise: in file mode, with an AI on your device: Chrome’s built-in AI or Gemma.
No upload: your audio stays on your device. How that works, and how to check it yourself, is explained in Private transcription without upload.

Who makes TranscriptBalance

TranscriptBalance is made by Jeremy Heindrichs from Bütgenbach in Belgium. He is a student and builds the app alongside his studies. Legally he is also the provider, as a sole proprietorship (natural person).

He also makes AudioBalance, a Windows app that keeps videos, calls and music at the same volume. The address, enterprise number and contact details are in the legal notice.

Why TranscriptBalance is free and stays free

TranscriptBalance is free, and it stays that way. That is possible because the computing happens on your device and not on servers that cost money every time someone uses them.

Your device does the work: transcription runs in your browser, so a minute of audio costs no server time.
No AI servers of its own: your browser downloads the AI models straight from Hugging Face and the runtime from jsDelivr. The pages themselves are static files on Cloudflare Pages; the only code that runs on the server is the redirect to your language and the anonymous counting (see below).
No subscription, no paid version: there is nothing to unlock and no quota that runs out.
No third-party ads: the only advertising is for the maker’s own product, a clearly labelled pointer to AudioBalance.

If you would like to support TranscriptBalance, have a look at AudioBalance.

Principles

Your audio never leaves your device. Files, microphone and shared audio are processed in your browser and are not uploaded anywhere. Of your recordings, only text is ever saved, never audio, and only in your browser.
No account, no cookies. There is no sign-up and no email address to give. Transcripts, summaries and settings are kept only in your browser’s storage.
Connections only when you click. The app connects to other providers only when you start a feature or click a link, never just because you opened the page.
No selling of data, no profiles. No data is sold and no profiles are created.

What does leave your device:

Model downloads: when you start a feature, your browser downloads the AI models from Hugging Face and the runtime from jsDelivr. These providers see your IP address and which file you fetch, but no content.
Gemini, only if you want it: in live mode you can connect Google’s Gemini with your own key (18 and over). Excerpts of the transcript and your questions then go straight from your browser to Google, not to TranscriptBalance. File mode never uses Gemini.
Anonymous counting: visits, the features used and clicks on links to AudioBalance are counted: never content, no cookies and no stored IP addresses. You can switch off the feature messages in the app under “About & licenses”; with Global Privacy Control or Do Not Track, your visit is not counted either.

All the details, including hosting at Cloudflare, calendar links and your browser’s own AI, are in the privacy policy.

The technology in brief

TranscriptBalance runs Whisper, OpenAI’s open speech recognition model, right in the browser. The computing is done with transformers.js and ONNX Runtime Web: through WebGPU on the graphics card or, where that is not available, through WebAssembly on the processor.

Transcription: four Whisper models (tiny, base, small and large-v3-turbo); the largest is only offered with WebGPU.
Live mode: small Whisper models on the processor, with a dedicated model for German, plus Silero VAD, which detects when someone is speaking.
Speaker detection: pyannote and WeSpeaker.
Summaries in file mode: Chrome’s built-in AI (Gemini Nano) or Google’s Gemma 3 1B, both on your device.
The rest: Mediabunny reads audio and video files, Svelte builds the interface, and the website is served by Cloudflare Pages.

TranscriptBalance would not exist without the free models and open-source libraries made by others. The Licences page lists who developed them and which licences apply. More on processing in the browser: Private transcription without upload.

The limits, stated plainly

A current browser: TranscriptBalance is fastest with WebGPU, for example in a current Chrome or Edge on a computer. Without WebGPU the processor does the work: slower, without the “Precise” model and without summaries by Gemma.
A one-time download: the first time, your browser downloads the model you picked, roughly 45 to 760 MB depending on the model and your device. After that it is kept in the browser’s storage.
Speed depends on your device: on weaker devices without WebGPU, a more accurate model can take as long as the recording itself, or longer. The app measures how fast your device is and suggests a model to match.
No account also means no cloud: results are kept only in this browser, with no syncing between devices. The tab has to stay open while the work is running.
Audio sharing only in Chrome and Edge: live mode can transcribe the audio of a tab or the screen only in Chrome or Edge on a computer; everywhere else it uses the microphone.
Mistakes happen: transcripts and AI summaries are generated automatically and can contain errors. Check names, numbers and quotes yourself.

The quickest way to see how well it runs on your device is to try it with a file of your own: no account, and your audio stays on your device.

Try TranscriptBalance

A short description to quote

For the press, directories and AI assistants, ready to use as it is:

TranscriptBalance (transcript-balance.com) is a free web app that converts audio and video files to text and transcribes lectures and meetings live. The Whisper speech recognition model runs in the browser on the user’s own device, using WebGPU or WebAssembly, so recordings are not uploaded and no account is needed. TranscriptBalance is developed by Jeremy Heindrichs, a student in Belgium, and has been publicly available since October 2026.

The key facts:

Name: TranscriptBalance (also written “Transcript Balance”)
Website: transcript-balance.com
What it is: a transcription web app that runs in the browser, with nothing to install
Price: free, no subscription, no limit on minutes or files
Account: none needed
Languages: 99 languages for transcription; interface in German, English, French and Dutch
Provider: Jeremy Heindrichs, Belgium
Public since: October 2026
From the same provider: AudioBalance (audio-balance.com), a Windows app

Contact

For questions, bug reports or press enquiries, email is the easiest way to reach Jeremy Heindrichs. The email address, phone number and postal address are in the legal notice.

Questions & answers: the most common questions about files, live mode, privacy and cost on one page.
Free transcription compared: TranscriptBalance next to other free options, with sources and dates.