This is an automatic speech recognition model optimized for Apple silicon chip devices. By converting to the MLX framework and quantizing to the FP8 format, it enables fast on-device speech transcription on Apple devices. The model is fine-tuned for verbatim accuracy and is particularly suitable for scenarios requiring high-precision transcription.

Audio ProcessingMlxOther

GigaAM-v3 is a basic model for Russian automatic speech recognition based on the Conformer architecture, with 220 - 240 million parameters. It is the third-generation model in the GigaAM series, pre-trained on 700,000 hours of Russian speech data using the HuBERT-CTC objective, providing state-of-the-art performance in a wide range of Russian ASR fields.

Audio ProcessingPytorchMultiple Languages

This is an English automatic speech recognition model developed by Applied Brain Research (ABR) based on the state space model (SSM). It has approximately 19 million parameters and can efficiently and accurately transcribe English speech into text. The model performs excellently on multiple benchmark datasets, with an average word error rate of only 10.61%. It supports real-time speech recognition and can run on low-cost hardware.

Audio ProcessingTransformersEnglish

This is a fine-tuned Whisper speech recognition model based on the unsloth/whisper-large-v3-turbo architecture. It uses the Unsloth and Huggingface TRL libraries to achieve 2x training acceleration, significantly improving training efficiency.

Audio ProcessingTransformersEnglish

This model is an automatic speech recognition model fine-tuned for Swahili based on the OpenAI Whisper-small model. It was trained on the Swahili part of the FLEURS-SLU dataset, significantly improving the transcription accuracy of Swahili. The word error rate was reduced by 68% compared to the base model.

Audio ProcessingTransformersOther

This is a multimodal model that supports multilingual processing, covering multiple fields such as natural language processing, code processing, and audio processing. It can perform various tasks such as automatic speech recognition, speech summarization, speech translation, and visual question answering.

MultimodalSafetensorsMultiple Languages

MERaLiON-SER-v1 is a parameter-efficient multilingual speech emotion recognition model that can simultaneously predict 7 discrete emotion categories and continuous emotion dimension values. This model is built on the Whisper-Medium encoder and adopts the LoRA adaptation technology, showing excellent performance in the multilingual speech emotion recognition task.

Audio ProcessingTransformersMultiple Languages

This model is an automatic speech recognition model fine-tuned on the Common Voice 17.0 dataset based on OpenAI Whisper-small. It is specifically optimized for the Twi language and can accurately recognize speech content.

Audio ProcessingTransformersOther

The Ming-flash-omni preview version is a multimodal large model built on the Ling-Flash-2.0 sparse mixture of experts (MoE) architecture, with a total of 100B parameters and only 6B parameters activated per token. This model is a comprehensive upgrade based on Ming-Omni, showing significant improvements in multimodal understanding and generation, especially in speech recognition, image generation, and segmentation editing.

MultimodalDiffusersEnglish

This is a Bambara speech recognition model fine-tuned based on the OpenAI Whisper Small model. It was trained on 24,738 Bambara audio samples and can accurately transcribe Bambara speech into text, with a character error rate as low as 21.69%.

Audio ProcessingSafetensorsOther

This model is an automatic speech recognition model fine-tuned on Latin based on OpenAI Whisper-small. It is trained using 67 hours of Latin audio data, with a character error rate (CER) of 20. It supports the task of converting Latin speech to text.

Audio ProcessingSafetensorsOther

MedWhisper Large ITA is a domain-adapted variant based on OpenAI Whisper Large v3 Turbo, specifically optimized for Italian medical speech recognition. This model is fine-tuned using LoRA technology on a carefully selected dataset of Italian outpatient specialty consultation recordings, significantly improving the recognition accuracy of medical terms and clinical expressions.

Audio ProcessingTransformersOther

This is an automatic speech recognition model based on the Hugging Face Transformers library, which can convert audio content into text. This model supports multiple languages and is suitable for scenarios such as real-time speech-to-text conversion and audio transcription.

Audio ProcessingTransformers

Borealis is the first automatic speech recognition (ASR) audio large language model for Russian, trained on approximately 7,000 hours of Russian audio data. This model supports the recognition of punctuation marks in audio. Its architecture is inspired by Voxtral but has been improved, and it performs excellently in multiple Russian ASR benchmark tests.

Audio ProcessingPytorchOther

This is a Russian automatic speech recognition model based on the OpenAI Whisper Small architecture, specifically trained on the Common Voice 17.0 dataset, capable of accurately converting Russian speech into text.

Audio ProcessingTransformersOther

A speech recognition model fine-tuned on the Common Voice 17.0 Swahili dataset based on OpenAI Whisper Large v2, designed specifically for the Swahili speech recognition task of the Tanzanian children's helpline. It has a significant improvement in Swahili recognition accuracy compared to the base model.

Audio ProcessingTransformersOther

SE-DiCoW is a target speaker multi-speaker automatic speech recognition model developed jointly by BUT Speech@FIT, JHU CLSP/HLTCOE, and CMU LTI. Based on Whisper large-v3-turbo, this model significantly improves the recognition accuracy in highly overlapped multi-speaker scenarios through a self-registration mechanism and improved data augmentation techniques.

Audio ProcessingTransformersOther

Moonshine Tiny is a lightweight Vietnamese automatic speech recognition model developed by Moonshine AI (formerly Useful Sensors). It has only 27M parameters and is specifically designed for resource-constrained platforms. It performs excellently on the Fleurs and Common Voice 17 datasets.

Audio ProcessingTransformersOther

An aviation traffic control speech recognition model fine-tuned based on the OpenAI Whisper Large v3 model, specifically optimized for ATC communication scenarios. It achieved a word error rate of 6.5% on the test set, significantly improving the accuracy of aviation communication transcription.

Audio ProcessingTransformersEnglish

parakeet-tdt-0.6b-v3 is a powerful multilingual automatic speech recognition model that supports multiple European languages such as English, Spanish, French, and German. It is based on the FastConformer-TDT architecture and trained on public datasets, providing an efficient solution for cross-lingual speech recognition.

Audio ProcessingNemoMultiple Languages