LuxASR Dictation

// Luxembourgish speech to text

About this project

A speech-to-text interface built for dictation, using the University of Luxembourg's LuxASR — whose own site is focused on audio transcription and offers only basic dictation. Pause and resume anywhere in the text, transcriptions insert at your cursor, and a period of silence can trigger the next transcription automatically. Fully keyboard operable.

Web app · built on the official LuxASR API

  • Speech to text
  • Luxembourgish
  • Accessibility
  • Web app
  • Free

Luxembourgish audio only, no file upload. This dictation-focused interface needs a microphone and a browser that supports MediaRecorder (Chrome, Firefox, Edge, Safari 14.1+).

The official LuxASR site is focused on transcribing audio files and offers many more features than this page, but its dictation mode is fairly basic. This page is a dedicated speech-to-text interface: you talk in bursts, correct yourself, and keep going. It's closer to the dictation feature built into a phone.

What this interface adds

  1. Continuous dictation. Suits long documents better than one-shot transcription.
  2. Cursor-aware. Each transcription is inserted wherever your cursor is in the text area.
  3. Auto-transcribe on silence. A pause in dication can trigger a transcription.

What it doesn't do

But the official interface does.

  • Other input languages: Luxembourgish only, no German, English, French or Portuguese input.
  • File upload: live microphone recording only.
  • Advanced output: plain text or JSON only, no SRT subtitles, Praat TextGrid, or word-level timestamps.
  • BAS TextGrid integration: not available here.

How your audio is handled

Recording happens in your browser. Each segment (whatever you've dictated since the last "Transcribe & continue" or "Finish") is sent to a small proxy on this site, which forwards it to the University of Luxembourg's LuxASR service and returns the text — nothing is written to disk here. Per LuxASR's own disclaimer, transcription runs on a dedicated University of Luxembourg server, audio is streamed rather than stored, and no data is used to train models or shared with third parties.

Accessibility

Every control is reachable and operable by keyboard alone, state changes (recording, paused, transcribing, errors) are announced to screen readers through live regions rather than colour alone, focus is always visible, and the pulsing status dot and countdown timer both respect a reduced-motion preference.

Use it

Ready to start recording

Recording options

Separate text by speaker
Format returned by LuxASR
Audio level below this counts as silence
Silence duration before auto-transcribe (0 to disable)
Audio level 0%

Edit this text area and position your cursor where you want the next transcription to be inserted.

How to use this tool

Recording controls

  • Start recording: begin capturing audio from your microphone.
  • Pause: temporarily stop recording without transcribing; press Start recording again to resume.
  • Transcribe & continue: transcribe the audio recorded so far, insert it at your cursor, and keep recording without a break.
  • Finish: stop recording and transcribe whatever's left.
  • Auto-transcribe: with a duration above 0 seconds, a silence that long finishes and transcribes automatically.

Keyboard

  • Tab moves between controls, Space or Enter activates the focused button.
  • Escape cancels an auto-transcribe countdown.

Further information

  • LuxASR Dictation does not store your recordings or transcriptions; everything lives in the text area in your browser until you copy it out.
  • LuxASR itself is developed by Peter Gilles, Léopold Hillah and Nina Hosseini-Kivanani at the University of Luxembourg; read more about the project.