### Open-source Danish speech recognition model coral-1-whisper-large - Accurate recognition with extremely low error rate

Coral 1 Whisper Large

Developed by Alvenir

A Danish speech recognition model trained on Whisper Large v3, fine-tuned on the CoRaL dataset, with excellent word error rate and character error rate performance.

Speech Recognition

Transformers

OtherOpen Source License:Openrail #Danish Speech Recognition #Low CER High Accuracy #Reading Scenario Optimization

Downloads 205

Release Time : 2/17/2025

Model Overview

This is an advanced Danish Automatic Speech Recognition (ASR) model trained by Alvenir, based on OpenAI's Whisper Large v3 architecture, fine-tuned on the CoRaL read-aloud dataset, significantly improving the accuracy of Danish speech recognition.

Model Features

High-Performance Danish Recognition

Achieves 4.3% CER and 10.4% WER on the CoRaL test set, outperforming other Danish ASR models.

Based on Whisper Architecture

Fine-tuned from OpenAI's Whisper Large v3 model, inheriting its powerful speech recognition capabilities.

Professional Dataset Training

Trained using the CoRaL professional Danish dataset, optimized for Danish language.

Model Capabilities

Danish Speech Recognition

High-Accuracy Transcription

Reading Speech Processing

Use Cases

Speech Transcription

Danish Meeting Minutes

Automatically transcribe Danish meeting recordings into text

Accuracy exceeds 90%

Educational Applications

Used for automatic subtitle generation of Danish learning materials

Low-error-rate transcription results

Voice Assistants

Danish Voice Interaction

Provide speech recognition support for Danish voice assistants

🚀 Whisper-Large v.3 trained on CoRaL release 1

This is a Danish state-of-the-art speech recognition model, trained by Alvenir, offering high - performance automatic speech recognition.

🚀 Quick Start

This Danish speech - recognition model is ready to use for automatic speech recognition tasks. You can start leveraging its capabilities right away.

✨ Features

State - of - the - art: It represents the current best in Danish speech recognition.
Trained on CoRaL: Utilizes the first release of CoRaL data for training, enhancing its performance.

📚 Documentation

Evaluation Results

Model	Number of parameters	CoRal CER	CoRal WER
[Alvenir/coral - 1 - whisper - large](https://huggingface.co/Alvenir/coral - 1 - whisper - large)	1540M	4.3% ± 0.2%	10.4% ± 0.3%
[alexandrainst/roest - 315m](https://huggingface.co/alexandrainst/roest - 315m)	315M	6.6% ± 0.2%	17.0% ± 0.4%
[mhenrichsen/hviske - v2](https://huggingface.co/syvai/hviske - v2)	1540M	4.7% ± 0.07%	11.8% ± 0.3%
[openai/whisper - large - v3](https://hf.co/openai/whisper - large - v3)	1540M	11.4% ± 0.3%	28.3% ± 0.6%

Results of more models and more datasets can be seen in the [model card for Røst - 315m](https://huggingface.co/alexandrainst/roest - 315m).

Model details

This is simply the [Whisper Large v.3 model](https://hf.co/openai/whisper - large - v3) trained on the first release of CoRaL data.

The model was trained for 30K steps using the configuration from the CoRaL repository by running:

python src/scripts/finetune_asr_model.py model=whisper-large max_steps=30000 model.learning_rate=1e-5

Information Table

Property	Details
Library Name	transformers
License	openrail
Datasets	alexandrainst/coral
Language	da
Metrics	wer, cer
Base Model	openai/whisper - large - v3
Pipeline Tag	automatic - speech - recognition

Model Index

Name: coral - 1 - whisper - large
- Results:
  - Task:
    - Type: automatic - speech - recognition
    - Name: Automatic Speech Recognition
  - Dataset:
    - Name: CoRal read - aloud
    - Type: alexandrainst/coral
    - Split: test
    - Args: read_aloud
  - Metrics:
    - Type: cer
    - Value: 4.3% ± 0.2%
    - Name: CER
    - Type: wer
    - Value: 10.4% ± 0.3%
    - Name: WER

📄 License

Note that the dataset used is licensed under a custom license, adapted from OpenRAIL - M, which allows commercial use with a few restrictions (speech synthesis and biometric identification). See [license](https://huggingface.co/Alvenir/coral - 1 - whisper - large/blob/main/LICENSE).

Creators and Funders

The CoRal project is funded by the Danish Innovation Fund and consists of the following partners:

We would like specifically thank Dan Saattrup Nielsen, Alexandra Institute for (among other things) the repository work and Simon Leminen Madsen, Alexandra Institute for modelling work.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご