K

Kb Whisper Large

Developed by KBLab
A Swedish speech recognition model based on the Whisper architecture released by the National Library of Sweden. The training data exceeds 50,000 hours, significantly reducing the word error rate.
Downloads 8,880
Release Time : 2/14/2025

Model Overview

A speech recognition model optimized for Swedish, based on the OpenAI Whisper architecture, and performs excellently on multiple Swedish datasets.

Model Features

Significantly reduce the word error rate
Compared with the original OpenAI model, it reduces the word error rate (WER) by an average of 47% in Swedish recognition.
Multi-format support
Provides model checkpoints in multiple formats, including Hugging Face, whisper.cpp (GGML), onnx, and ctranslate2.
Multiple transcription styles
Provides three transcription style versions: subtitle version (concise), standard version (default), and strict version (verbatim).
Large-scale training data
Trained on over 50,000 hours of Swedish speech data, with training conducted in two quality stages.

Model Capabilities

Swedish speech recognition
Speech transcription with timestamps
Multi-format inference support
Batch speech transcription

Use Cases

Speech transcription
Meeting record transcription
Convert Swedish meeting recordings into text records.
High-accuracy transcribed text
Subtitle generation
Generate subtitles for Swedish video content.
Subtitle files with timestamps
Speech analysis
Speech content analysis
Analyze Swedish speech content for subsequent processing.
Structured text data
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase