# High-precision ASR

Quantum_STT is an advanced automatic speech recognition (ASR) and speech translation model, trained with large-scale weak supervision, supporting multiple languages and tasks.

GigaAM-v2-RNNT is a Russian automatic speech recognition (ASR) model based on the RNNT architecture, suitable for speech-to-text tasks.

A Persian automatic speech recognition model fine-tuned based on Whisper Large v3, trained on the Common Voice 17 dataset, which contains over 250,000 Persian audio samples, significantly improving recognition accuracy and robustness.

NVIDIA NeMo Canary Flash is a family of multilingual multitask models that achieves state-of-the-art performance across multiple speech benchmarks. Supports automatic speech recognition and translation tasks in four languages.

A Korean automatic speech recognition (ASR) and speech translation (AST) model fine-tuned based on microsoft/Phi-4-multimodal-instruct, demonstrating excellent performance on the zeroth-korean and fleurs datasets.

A fine-tuned version of OpenAI Whisper Large v3 model specifically for Hebrew language audio transcription tasks

An Arabic automatic speech recognition model fine-tuned on the QASR dataset, specifically adapted for dialectal variants

Whisper-V3-Turbo model optimized for Vietnamese automatic speech recognition (ASR) tasks, fine-tuned using multiple Vietnamese datasets

An English automatic speech recognition model pre-trained on the GigaSpeech dataset, supporting both streaming and non-streaming transcription

An English speech recognition model jointly developed by NVIDIA NeMo and Suno.ai, supporting punctuation and capitalization conversion, based on FastConformer-TDT-CTC architecture

This is an automatic speech recognition model optimized for Catalan, fine-tuned based on OpenAI's Whisper-large-v3 and developed by the Barcelona Supercomputing Center.

Parakeet TDT-CTC 0.6B is an automatic speech recognition (ASR) model capable of transcribing Japanese speech with punctuation, developed by the NVIDIA NeMo team.

This is an end-to-end automatic speech recognition system pre-trained on the LibriSpeech dataset, supporting both streaming and non-streaming modes, suitable for English speech recognition.

Canary-1B is a multilingual multi-task model developed by NVIDIA NeMo, supporting automatic speech recognition and speech translation tasks in English, German, French, and Spanish.

Norwegian automatic speech recognition model developed based on OpenAI Whisper, with additional training for lowercase, punctuation-free verbatim transcription

Whisper is an advanced automatic speech recognition (ASR) and speech translation model proposed by OpenAI, trained on over 5 million hours of labeled data, with strong cross-dataset and cross-domain generalization capabilities.

NVIDIA FastConformer-Hybrid Large (ua) is a hybrid model for Ukrainian speech recognition, which combines the training of two loss functions, Transducer and CTC, with approximately 115 million parameters.

A SpeechT5 automatic speech recognition model fine-tuned on the LibriSpeech dataset, supporting speech-to-text conversion.

A Mongolian speech recognition model fine-tuned on Mongolian datasets based on OpenAI's whisper-large-v2 model, supporting automatic speech recognition tasks in Mongolian.

Fine-tuned on an enhanced Thai dataset based on openai/whisper-medium for Thai automatic speech recognition

A speech recognition model fine-tuned on the Zeroth Korean dataset based on OpenAI Whisper Medium model, with a word error rate of 3.64%

A Chinese speech recognition model fine-tuned based on microsoft/wavlm-large, supporting Simplified Chinese, trained using the Common Voice 7.0 (zh-CN) dataset.

An Italian automatic speech recognition model fine-tuned from the facebook/wav2vec2-large-100k-voxpopuli model, trained using the Common Voice 7.0 Italian dataset.

An Italian automatic speech recognition model fine-tuned based on facebook/wav2vec2-large-lv60, trained using the Common Voice 7.0 Italian dataset.

A Japanese automatic speech recognition model fine-tuned using the training set of Common Voice 7.0 (Japanese version), based on the facebook/wav2vec2-large-it-voxpopuli model.

This model is an automatic speech recognition model fine-tuned on the GARY109/AI_LIGHT_DANCE - ONSET-SINGING2 dataset based on wav2vec2-large-xlsr-53, primarily used for singing voice recognition tasks.

This model is a fine-tuned speech recognition model based on facebook/wav2vec2-xls-r-300m on the common_voice dataset, achieving a low word error rate on the evaluation set.

A Dutch automatic speech recognition (ASR) model developed by Facebook based on the Wav2Vec 2.0 architecture, fine-tuned using the XLSR-53 multilingual pretrained model

This is an automatic speech recognition (ASR) model fine-tuned on Maltese speech datasets based on the facebook/wav2vec2-xls-r-300m model.

Automatic speech recognition model based on the WavLM architecture, fine-tuned on the LibriSpeech CLEAN dataset (100 hours)

Large-scale Italian automatic speech recognition model based on the Wav2Vec2 architecture, fine-tuned on the Common Voice dataset, released by Facebook

This is a 300M-parameter Bengali speech recognition model based on the wav2vec2 XLS-R architecture, specifically optimized for command recognition tasks.

A speech recognition model based on Facebook's Wav2Vec2 architecture, pre-trained with 10K unlabeled Czech data from the VoxPopuli corpus and fine-tuned on Czech transcription data.

An automatic speech recognition model fine-tuned on the Welsh Common Voice dataset based on facebook/wav2vec2-large-xlsr-53, achieving a test WER of 29.4%.

Vietnamese automatic speech recognition model fine-tuned on 40 hours of FPT Open Speech Dataset (FOSD) and Common Voice 7.0 dataset based on facebook/wav2vec2-xls-r-1b

An English automatic speech recognition (ASR) model fine-tuned based on microsoft/wavlm-base, trained on the english_ASR - CLEAN dataset with a word error rate (WER) of 0.0773.

A speech recognition model based on Facebook's Wav2Vec2 architecture, pretrained on 10K unlabeled French data from the VoxPopuli corpus and fine-tuned on French transcription data.

This model is an automatic speech recognition (ASR) model fine-tuned on the Turkish COMMON_VOICE dataset based on cahya/wav2vec2-base-turkish-artificial-cv

A large-scale Portuguese automatic speech recognition (ASR) model developed by Facebook based on the Wav2Vec 2.0 architecture, supporting Portuguese speech-to-text tasks.

Polish automatic speech recognition model developed by Facebook, based on Wav2Vec2 architecture and XLSR-53 multilingual pretrained model