Murmel in De Telegraaf: building instead of talking

Dutch national newspaper De Telegraaf interviewed me about Murmel. The article appeared in the financial section under the headline "Entrepreneur builds Dutch AI alternative: building instead of talking", and covers why I built a speech model specifically for Dutch, and why it runs entirely on servers in the Netherlands.

Read the article at De Telegraaf → (Dutch)

Building instead of talking

There is a lot of talk in the Netherlands and Europe about digital independence: relying less on large foreign tech companies, building more technology of our own. Murmel is meant as a working answer to that debate. Not a plan or a position paper, but a speech recognition model that already exists: trained on Dutch speech, running on Dutch servers, and outperforming every open-source alternative for Dutch audio on public benchmarks.

Why a dedicated Dutch model?

International speech models are trained to handle dozens of languages at once. That breadth comes at the cost of depth: Dutch accents, dialects, and domain jargon are recognised less reliably. Murmel is trained specifically on thousands of hours of Dutch audio, from regional accents (the Haagse h, the zachte g) to fast speakers and meeting jargon.

Audio stays in the Netherlands

Everything runs on infrastructure in the Netherlands. Audio never leaves the country, which matters for organisations working with sensitive recordings (healthcare, government, media, and research). For that reason Murmel is also listed in the Tech Sovereignty Catalogue, the European register of digital solutions contributing to technological independence.

Try it yourself

Try Murmel two weeks for free, or buy prepaid credits from €5 without a subscription. Or get in touch to discuss a larger deployment.