Our own product
Hushscript
Hushscript is our own product for people who work with sensitive recordings, and this is the case study of how we built it. It pairs a Vue 3 app with a Rust backend compiled to WebAssembly and run at the edge, a static Astro site, an MCP server and a REST API, all designed, built and run by us.
What is Hushscript?
Hushscript turns recordings into clean, speaker-labeled transcripts for people who work with sensitive material: interviews, meetings, and medical and legal calls. It is used from the browser on desktop, tablet or phone, with nothing to install and no local AI models to set up.
Hushscript is our own product. Vasilii Dementev founded Hushscript and Lotus Web Agency; we designed and built it.
What shaped the way it had to work?
Compatible video is prepared in the browser, so only the audio the service needs is sent for processing, and that audio is deleted as soon as the transcript is ready.
It also had to work with nothing to install, no local AI setup, no subscription and no meeting bot sitting in every call, on a phone as well as a desktop.
What did we build to meet that?
- A Rust backend compiled to WebAssemblyIt runs on an edge runtime close to the user, with near-instant startup, and Rust’s memory safety keeps the code that handles recordings fast and predictable.
- Typed contracts between app and backendThe Vue 3 app is TypeScript throughout and shares typed contracts with the backend, so a changed endpoint fails type-checking before anything ships.
- A static Astro site and a Vue 3 appThe public site is pre-rendered, with no framework to boot before content appears. The single-page app is built with Vue 3, Pinia and TypeScript.
- In-browser media processing with ffmpeg.wasmAudio extraction, conversion and compression run in the browser, so the free tools never upload your files.
- An MCP server and a REST APIA remote MCP server with OAuth sign-in and 50+ tools lets AI assistants and agents that support MCP upload, transcribe, edit and export inside an account. The REST developer API comes with an OpenAPI document.
- Pay-as-you-go billing and passwordless sign-inPayments run through a payment provider, sign-in uses magic links, and bot protection is built in: no accounts to babysit and no monthly plan.
What has the work delivered?
- Hushscript is live at hushscript.com, a public site in 12 languages, with a Chrome extension alongside the web app.
~99
languages auto-detected
21
export formats
0
videos uploaded — audio only
No
subscription — pay per minute
What is Hushscript built on?
Who runs Hushscript today?
We design, build and operate Hushscript ourselves, from the app and the API to the infrastructure underneath.
The public site uses the same static-first approach as the one you are reading now.
Where can you try Hushscript?
A few good places to start:
Hushscript audio to text
Turn MP3, M4A, WAV, and other audio files into clean, speaker-labeled transcripts.
Hushscript video to text
Compatible video is prepared in your browser and only the audio it needs is uploaded.
Hushscript M4A to MP3 converter
One of the free in-browser tools — convert audio with nothing leaving your browser.
Why Hushscript
How it compares to subscription and local-AI transcription, and what "private by design" means.
Hushscript MCP server
How to connect AI assistants and agents that support MCP to a Hushscript account, and what the tools cover.
Hushscript developer API
The REST API for scripts and agents: tokens, uploads, jobs and transcripts, with an OpenAPI document.
Which services does this project draw on?
Try it yourself
Hushscript gives you 30 free minutes to start — no watermark, no monthly plan.