# Low WER

Whosper-large-v2 is a cutting-edge speech recognition model specifically designed for Wolof, the primary language of Senegal. Built upon OpenAI's Whisper-large-v2, it significantly improves Word Error Rate (WER) and Character Error Rate (CER).

A Hindi-Hinglish mixed speech recognition model optimized based on the Whisper architecture, specifically designed for Indian accents and noisy environments

Whisper-medium model optimized for Vietnamese speech recognition tasks, fine-tuned on 1308 hours of Vietnamese data

Parakeet CTC 0.6B is an automatic speech recognition model jointly developed by NVIDIA NeMo and Suno.ai, based on the FastConformer architecture with approximately 600 million parameters, supporting English speech transcription.

Parakeet RNNT 0.6B is an automatic speech recognition model jointly developed by NVIDIA NeMo and Suno.ai, based on the FastConformer architecture with approximately 600 million parameters, specifically designed for transcribing English speech into text.

Parakeet CTC 1.1B is an automatic speech recognition model jointly developed by NVIDIA NeMo and Suno.ai, based on the FastConformer architecture with approximately 1.1 billion parameters, supporting English speech transcription.

A French automatic speech recognition model fine-tuned based on OpenAI Whisper-large-v3, supporting case sensitivity, punctuation, and number prediction

A Whisper medium speech recognition model fine-tuned on the CommonVoice Arabic dataset, developed by the SpeechBrain team

The NVIDIA FastConformer-Transducer is a high-performance model for English automatic speech recognition (ASR), utilizing an optimized FastConformer architecture and Transducer decoder with approximately 618 million parameters.

NVIDIA FastConformer-CTC XLarge is an Automatic Speech Recognition (ASR) model with approximately 600 million parameters, designed specifically for English speech transcription and trained using the FastConformer architecture and CTC loss.

A French automatic speech recognition model fine-tuned based on openai/whisper-small, trained on the Common Voice 11.0 French dataset, supporting case sensitivity and punctuation prediction.

This is a large Conformer-Transducer model for Kinyarwanda speech recognition, which can transcribe speech into lowercase Latin letters, supporting spaces and apostrophes.

This is a large Conformer-Transducer model for Spanish automatic speech recognition, with approximately 120 million parameters, trained on 1340 hours of Spanish speech data.

This is a large Conformer-Transducer model for German automatic speech recognition, with approximately 120 million parameters, supporting the transcription of German speech into text.

This is a large-scale Conformer-CTC model for German automatic speech recognition, trained and optimized by NVIDIA on thousands of hours of German speech data.

A Chinese speech recognition model fine-tuned on the AISHELL-1 dataset based on facebook/wav2vec2-large-xlsr-53, supporting Chinese speech recognition tasks.

This is an automatic speech recognition (ASR) model fine-tuned on the German Common Voice 9.0 dataset, based on Facebook's wav2vec2-large-xlsr-53 model.

Vietnamese speech recognition model based on Wav2vec2, fine-tuned on 160 hours of Vietnamese speech data

Danish speech recognition model fine-tuned on the NST dataset, supporting 16kHz sampled audio input

This is an automatic speech recognition model fine-tuned on the Turkish Common Voice dataset based on the facebook/wav2vec2-large-xlsr-53 model, achieving a test WER of 21.13%.

Speech recognition model optimized for Brazilian Portuguese, trained with 453 hours of audio from 7 public datasets

A Sundanese speech recognition model fine-tuned from facebook/wav2vec2-large-xlsr-53, trained on high-quality TTS data from OpenSLR

wav2vec 2.0 speech recognition model trained on the CommonVoice French dataset, using CTC/Attention architecture without requiring a language model

A Wav2vec 2.0 speech recognition model fine-tuned on Brazilian Portuguese datasets, supporting automatic speech recognition tasks for Brazilian Portuguese.

A speech recognition model fine-tuned for Esperanto using the Common Voice dataset, based on the facebook/wav2vec2-large-xlsr-53 model.

This is a Wav2vec2 model optimized for Brazilian Portuguese, trained on multiple open datasets for automatic speech recognition tasks.

This is a Wav2vec 2.0 model fine-tuned for Brazilian Portuguese, trained using multiple open Brazilian Portuguese datasets including Common Voice, MLS, CETUC, etc.

This model is a fine-tuned speech recognition model based on facebook/wav2vec2-base-960h on the TED-LIUM dataset, achieving a word error rate of 20.33% on the evaluation set.

This is a Wav2vec 2.0 model fine-tuned for Brazilian Portuguese, trained on multiple Brazilian Portuguese datasets, achieving a WER of 13.6 on the Common Voice test set.

A Japanese speech recognition model fine-tuned based on facebook/wav2vec2-large-xlsr-53, supporting hiragana output

A speech recognition model fine-tuned on the Spanish Common Voice dataset based on Facebook's wav2vec2-large-xlsr-53 model, with a test WER of 10.50%.

This is an Esperanto speech recognition model fine-tuned from Facebook's wav2vec2-large-xlsr-53 model, trained using the Common Voice dataset.

This is an automatic speech recognition model based on the XLS-R architecture, specifically optimized for Dutch and Flemish, incorporating a 5-gram language model to improve recognition accuracy.

This model is an automatic speech recognition model fine-tuned on the Galician dataset based on facebook/wav2vec2-xls-r-300m, achieving a WER of 11.31% on the Common Voice 8.0 test set.