Why audio-to-text so often disappoints
Search for "audio to text" and you will find dozens of tools. The experience is often the same: the first result looks decent, but compare the transcript to the recording and you will find mangled names, invented sentences where there was background noise, and entire passages missing. The cause is usually simple: most transcription models are trained primarily on English. Dutch is an afterthought, and accents, colloquialisms and jargon suffer.
Murmel runs on a model trained specifically for Dutch speech. In a public benchmark on 8.9 hours of Dutch parliamentary audio it makes fewer errors than open-source alternatives such as Whisper in every speaker group tested, especially for speakers with regional accents. The full methodology and numbers are online so you can verify them yourself.
Where is your audio, actually?
The second problem is invisible: many free and cheap tools process your recording on American servers. Your audio then falls under US law, and what happens to it afterwards (retention, model training) is often unclear. For a private file that is uncomfortable; for an interview, meeting or consultation it is a real risk.
With Murmel everything stays in the Netherlands:
- Processing on servers in the Netherlands; your audio never leaves the EU.
- Recordings and transcripts are automatically deleted after 7 days.
- Your data is not used for AI training unless you explicitly opt in.
How it works
- Upload your file. mp3, m4a, wav or video (mp4): it all works, up to files several hours long.
- Murmel transcribes. The text is ready within minutes, with speaker labels if you want them.
- Correct and export. In the editor you replay the audio next to the text, adjust where needed and export to Word or subtitles.
Trial or prepaid credits: no strings attached
You can use Murmel in two ways. Try it for 14 days for €0.01 with the full Basic plan (then €9 per month excl. VAT, cancel monthly). Or buy prepaid credits from €5 without a subscription: €1 equals 50 minutes of audio (excl. VAT) and credits stay valid for 6 months. For a one-off job, credits are usually the logical choice.
Frequently asked questions
Which file formats can I upload?
All common audio and video formats, including mp3, m4a, wav and mp4. Long recordings of several hours are no problem.
How accurate is the transcription?
Murmel uses a model trained specifically for Dutch speech. In a public benchmark on Dutch parliamentary audio it makes fewer errors than open-source alternatives such as Whisper in every speaker group tested. The exact numbers and methodology are on the benchmark page.
Where is my audio processed?
On servers in the Netherlands. Your audio never leaves the EU, is automatically deleted after 7 days, and is not used for AI training unless you explicitly opt in.
What does converting audio to text cost?
Prepaid credits cost €1 per 50 minutes of audio (excl. VAT), available from €5 and valid for 6 months. Or try 14 days for €0.01 and then €9 per month (excl. VAT), cancel monthly.
How fast is it?
Most files are transcribed within minutes. You get a notification as soon as the transcript is ready, so there is no need to wait around.