Murmel Blog
News, updates, and insights on our Dutch speech recognition and transcription technology.
Murmel v3: faster, more efficient and better
A new model built from the ground up for Dutch: 38% fewer errors, an hour of audio in under a minute, speaker recognition built in and Dutch and English mixed.
Read more →Murmel Teams: one pool of minutes for your whole team
Buy a plan for several seats, invite colleagues by email address and share a single pool of minutes. Because in every team one person does the bulk of the transcribing, and a personal quota is exactly what blocks them.
Read more →New: Murmel automatically transcribes your meetings
Paste a Google Meet or Microsoft Teams link and Murmel joins as a participant, records the meeting, and delivers the transcript afterwards. No record button, no hunting for the file, no upload.
Read more →Murmel in De Telegraaf: building instead of talking
Dutch newspaper De Telegraaf interviewed the maker of Murmel about the Dutch AI alternative: a speech model trained on Dutch speech, running entirely on servers in the Netherlands.
Read more →Murmel 2.1: better transcriptions and no more limit on file duration
Improved transcriptions based on user feedback, and the end of the file duration limit: long recordings of many hours are now automatically processed in parts. Upload and done, Murmel handles the rest.
Read more →Murmel v2 available now: 30% fewer errors in Dutch speech-to-text
Murmel v2 makes about 30% fewer errors than v1 and ranks third on public Dutch benchmarks, ahead of OpenAI, Mistral, NVIDIA, Cohere and Qwen. Only two very well-funded frontier systems are still ahead.
Read more →New: upload video and export it with subtitles
Upload an mp4 or webm, have it transcribed at up to 35× faster than realtime, and select a piece of text to export exactly that fragment as video with subtitles.
Read more →Twice as many minutes, and prepaid credits 5× cheaper
All paid subscriptions get double the included minutes starting today — same price. Prepaid credits drop from €0.10 to €0.02 per minute. Scale benefits passed straight on to users.
Read more →Murmel on BNR: Dutch speech model outperforms big tech
Dutch business radio BNR featured Murmel: a conversation about Dutch speech recognition, accents and jargon, and the basement in Zaandam where the GPUs run.
Read more →Murmel listed in the Tech Sovereignty Catalogue
The European DIGITAL SME Alliance has added Murmel to its European register of sovereign digital solutions — confirmation that our setup (Dutch hosting, no audio leaving the country) meets the criteria for technological independence.
Read more →How Murmel performs on speakers from different provinces
Methodology of the Murmel ASR benchmark on Dutch parliamentary debate audio: which audio, which models, how the metadata was collected, and the limitations of the numbers.
Read more →New: Custom Vocabulary for flawless transcriptions with jargon
However advanced AI is, jargon remains a challenge. That is why we are introducing Custom Vocabulary: prime Murmel with specific names and terminology.
Read more →Meet Murmeltje: our new logo is a marmot!
Murmel has a new face: Murmeltje, a happy marmot ready to help you transcribe. Celebrate with code MURMELTJE40 for 40% off your first month!
Read more →Prepaid credits: transcribe without a subscription
Not everyone transcribes every week. You can now buy prepaid credits: €0.02 per minute, pick your own amount from €5, valid for 6 months, no subscription.
Read more →First test results: Better than Mistral on Dutch
Read about the first user experiences with Murmel. The model outperformed Mistral on a Dutch benchmark and perfectly transcribed a mumbling speaker.
Read more →Speaker diarisation, accessibility, and sustainable infrastructure
After a week of testing with podcast makers, cultural organisations, and civil servants: speaker diarisation is live, the interface is more accessible, and Murmel processes audio at 37× realtime.
Read more →Murmel is coming: the platform and API are launching soon
Discover the new platform built around our Dutch speech-to-text model. 100% hosted in the Netherlands, including a user-friendly interface and robust REST API.
Read more →Introducing Murmel: a Dutch speech-to-text model built in the Netherlands
Existing open-source models struggle with context and accents. That's why we trained our own model, optimized for the Dutch language (17.3% WER).
Read more →