Skip to content
Introducing Klang Pianissimo for Danish

Pianissimo.

The most accurate Danish speech-to-text.

Klang Pianissimo is an open speech-to-text model for Danish, fine-tuned from NVIDIA’s Parakeet v3. It sets a new standard for Danish, with the highest accuracy on real, everyday speech.

Danish speech-to-text model trained on real speech.

Pianissimo makes fewer errors on real Danish speech than any other model we tested, open or closed: 10% fewer than Edda, 11% fewer than Saga and 4% fewer than ElevenLabs Scribe v2.

Try Pianissimo

Try transcribing H.C. Andersen, or try the model on your own audio.

Connecting to the model…

H.C. Andersen

Den lille pige med svovlstikkerne og Fyrtøjet

0:0023:02
Or try your own audio

Drop a file onto the player. Up to 90 minutes. Audio is sent to Klang for transcription.

The best Danish speech recognition model

Open source and fast

84%85%86%87%88%89%Accuracy85.1Google Gemini 3.5 Transcribe85.4Mistral Voxtral Mini Transcribe 385.7edda-v0.285.7svale-110M85.8whisper-small-danish85.8OpenAI GPT Transcribe85.9edda-v0.2-large85.9svale-600M86.0whisper-large-v3-turbo-danish86.0Syv Transcribe86.1ekko-v1-tiny86.2Mistral Voxtral Small Transcribe 386.3edda-v0.186.5hviske-v686.7saga-2-m86.8brage-v186.9edda-v0.2-duo87.7ElevenLabs Scribe v288.2Klang Pianissimo
Accuracy = 100 − WER (word error rate). Average of three Danish test sets: parliament speeches, spontaneous speech and TV subtitles.

60 hours of audio in one minute.

1 second

to transcribe an hour of audio

Audio transcribed in one minute of processing
One tile = one hour of audio

An open ecosystem
for Nordic speech.

Most large speech models are trained mainly on English, and the Nordic languages fall behind. We think that's best solved together, in the open. That's why Pianissimo is released with open weights, so anyone can build on it and share what they make. This is our way of contributing to the Nordic speech ecosystem.

Get Pianissimo.

Pianissimo is an open model, built for sovereignty. Download the model weights and run it on your own infrastructure. You choose where your audio is processed.

Hugging Face

Run on your own infrastructure.

Download the weights, adapt the model and build on it. You choose the hardware and where your audio is processed.

You choose where it runs.

European AI needs models that others can build on. Open weights let you run Pianissimo locally, choose a European provider and switch as your needs change.

Built with Pianissimo.

Pianissimo is open under CC BY 4.0 and runs on your own computer. Build something with it and submit your project to be featured here.

  • Built on Pianissimo for Swedish
  • Works with Pianissimo for Danish
Open source

Pianissimo CLI

Transcribe files, folders and podcasts locally.

What are you building?

We’d love to see your project and share more ways to use Pianissimo.

Share your project
Klang

AI for conversations that matter.

Klang is a Nordic AI speech lab building a perfect memory of all your conversations - with transcriptions, summaries, and insights. Built in Europe, with trust at its core.