noScribe – Free Interview Transcription with AI#

This is the official project site of the open-source software noScribe (GitHub). noScribe is a desktop app for transcribing interviews and other sensitive audio data – privacy-friendly, without cloud, with high-quality local AI models.

Warning: Someone is trying to sell paid transcription services via the domain noscribe(dot)ai. I have nothing to do with that. The real noScribe is free and will always remain so :)

What is noScribe?#

• Free and open source (GPL-3.0)
• No cloud, everything stays on your own computer
• High-quality transcripts based on Whisper by OpenAI and Pyannote by Hervé Bredin
• Speaker recognition and support for around 60 languages
• available for Windows, macOS, and Linux
• Comfortable editor for reviewing and correcting transcripts

Originally developed for qualitative social research, noScribe is now used wherever particularly sensitive audio data is handled - in journalism, medicine, companies, government agencies and NGOs. noScribe is recommended by universities, research platforms, the German computer magazine c’t, and in professional publications for journalists.

(The screenshot shows an interview with the exiled Russian sociologist Natalia Savelyeva that I conducted in May 2022.)

How do I use noScribe?#

Input and output

noScribe can process almost any audio or video format. By default it creates an HTML file that can be opened directly in MS Word, with line numbers and formatting. It can also produce plain text or a subtitle file in WebVTT format. WebVTT is mainly intended for further processing with EXMARaLDA, specialized software for complex transcription systems.

Processing

Because transcription happens entirely locally on your own computer, it needs some time: One hour of interview takes about 1-3 hours of transcription time, on older machines even longer. But you can queue multiple jobs, for example overnight. An NVIDIA graphics card (CUDA) can significantly speed up transcription. Since the download also includes the AI models, the installer is quite large (about 3 GB).

Special features

noScribe includes several features specifically for qualitative social research: pauses or overlapping speech can be marked and you can insert timestamps ([00:01:38]). You can also decide whether filler words and disfluencies like “um” should be transcribed or not.

Languages and correction

The AI model behind noScribe supports many languages and can handle dialects quite well, too. In general, a good audio recording with as little background noise as possible saves a lot of time in post-processing. But review and correction are always necessary. For this, a dedicated editor is included that lets you align audio and text easily.

More info:

• The noScribe usage guide and FAQ.
• Review and tutorial video (external)

Who is behind noScribe?#

Dr. Kai Dröge
Sociologist with a computer science background, researcher and lecturer

• My homepage at the Lucerne University of Applied Sciences and Arts (Switzerland)
• and at the Institute for Social Research, Frankfurt am Main (Germany).
• My YouTube channel “Qualitativ Forschen” (mostly German)

Many other people have contributed to noScribe. My special thanks go to Philipp Schneider (gernophil), who has long maintained the macOS versions, and Rudolf Siegel (mutlusun), who contributed many improvements.

Donations#

noScribe is free and always will be. But development causes real expenses, for example for this website, test hardware, or an Apple developer license. If you want to support the project:

Why the name “noScribe”?#

The Urban Dictionary defines a scribe as “a person whose entire miserable existence has been reduced to academic grunge and pain”. I hope this software will make our academic life a little less painful and grungy, hence the name noScribe :)

Citation (APA style)#

Dröge, K. (2026). noScribe. AI-powered Audio Transcription (Version XXX) [Computer software]. https://noscribe.de