Faster Whisper Small.en Alternatives

Systran/faster-whisper-small is an open-source speech recognition model converted from OpenAI's Whisper, optimized for use with CTranslate2. It enables fast and accurate transcription of audio files into text for developers and researchers.

This is a conversion of OpenAI's Whisper large-v2 model to the CTranslate2 format for faster inference. It supports automatic speech recognition in 99 languages and can be used through the faster-whisper Python package. The model is provided in FP16 precision and is suitable for local transcription tasks.

This repository provides a CTranslate2 conversion of the distil-whisper/distil-medium.en model for fast English speech recognition. It is designed for use with faster-whisper and other CTranslate2-based projects, offering significantly faster inference than the original PyTorch model while maintaining accuracy.

This repository provides the Whisper-medium model converted to CTranslate2 format for significantly faster inference. It can be used via the faster-whisper Python package. The model supports transcription across many languages and is intended for developers who require efficient, local speech recognition.

Faster Whisper Base.en is a converted version of the Whisper base.en model provided in CTranslate2 format for English automatic speech recognition. It addresses the need for accelerated inference on audio transcription tasks compared to the original PyTorch implementation. The model originates from the openai/whisper-base.en checkpoint and is hosted on Hugging Face by Systran. It can be loaded and run directly in CTranslate2 or in projects built on that library, such as the faster-whisper Python package. An example loads the model with the identifier "base.en", transcribes an audio file such as audio.mp3, and iterates over resulting segments that each contain start time, end time, and recognized text. The conversion process used the ct2-transformers-converter tool with the command that specified the source model, output directory, copied the tokenizer.json file, and applied float16 quantization. Model weights are stored in FP16, though the compute_type option in CTranslate2 allows changing the precision at load time. The repository carries an MIT license. It is intended for developers integrating English speech-to-text capabilities into applications that require efficient processing of audio files or streams. Delivery occurs as downloadable model files through the Hugging Face model hub, compatible with CTranslate2-based runtimes rather than the native PyTorch version of Whisper.

faster-whisper-large-v3 is an open-source automatic speech recognition model converted for use with CTranslate2. It supports 100 languages and is suitable for developers and researchers building local or embedded speech-to-text solutions.

faster-whisper-small-int8 is an optimized, INT8-quantized version of the OpenAI Whisper small model for speech-to-text. Created by the Rhasspy team, it is designed for efficient local inference with significantly reduced memory usage and faster performance while maintaining good transcription accuracy. It is available on Hugging Face for use with Transformers or faster-whisper libraries.

This repository provides the Whisper large-v3-turbo model converted to CTranslate2 format for use with faster-whisper and other CTranslate2-based projects. It supports transcription of audio in 100 languages with significantly improved inference speed. The model is intended for developers integrating fast, local speech recognition into applications.

whisper-tiny.en is a highly optimized English-only version of OpenAI's Whisper model converted by Xenova for ONNX Runtime. It enables fast, local speech recognition in web and Node.js environments without requiring large model downloads.

faster-whisper-tiny-int8 is an open-source, quantized speech-to-text model optimized for fast and efficient transcription. It is suitable for developers who need lightweight and accurate speech recognition in their applications, especially on limited hardware.

whisper-small is an open-source automatic speech recognition model developed by OpenAI. It supports multilingual audio transcription and is suitable for developers and researchers building speech-to-text applications. The model can be fine-tuned and deployed locally or via API for various audio processing tasks.

This is a converted version of OpenAI's Whisper-small model exported to ONNX format for use with Transformers.js and other ONNX runtimes. It performs automatic speech recognition, converting spoken audio into text in multiple languages. The model is designed for integration into web applications and environments where native PyTorch is not suitable.

The faster-whisper-large-v3-turbo model is a conversion of OpenAI's Whisper large-v3-turbo to the CTranslate2 format hosted on Hugging Face. It enables speech-to-text transcription through the faster-whisper library and other CTranslate2-based projects. The repository provides the converted model weights along with example code for loading and running transcription. A typical usage imports the WhisperModel class from faster_whisper, instantiates the model with the repository identifier, calls the transcribe method on an audio file such as audio.mp3, and iterates over the resulting segments to access start time, end time, and text. The weights are stored in FP16 precision, which can be adjusted at load time through the compute_type option in CTranslate2. The conversion from the original OpenAI model was performed using the ct2-transformers-converter command-line tool. It is called automatically by the Mobius Labs fork of faster-whisper. The model carries an MIT license and is published by Dropbox Inc. No pricing information is stated because the model is distributed as an open artifact on the Hugging Face platform.

openai/whisper-tiny.en is the smallest English-only variant of OpenAI's Whisper automatic speech recognition model. It converts English audio into text with high accuracy while using minimal computational resources. Hosted on Hugging Face, it is widely used for testing, research, and lightweight transcription applications.

This is a fine-tuned version of OpenAI's Whisper Small model specialized for Cantonese (zh-HK). It performs automatic speech recognition on Cantonese audio. The model is hosted on Hugging Face and follows the standard Whisper pipeline for inference.

faster-whisper-base-int8 is an open-source, quantized speech-to-text model designed for fast and efficient inference. It enables developers to transcribe audio to text with reduced resource requirements, making it suitable for real-time and embedded applications.

Xenova/whisper-tiny provides a browser-compatible version of OpenAI's Whisper tiny model using ONNX and Transformers.js. It enables client-side automatic speech recognition directly in web applications without sending audio to external servers. The model is optimized for low-resource environments and can be used via npm in any JavaScript or TypeScript project.

This is a small-sized Whisper model fine-tuned specifically for the Tamil language. It provides speech-to-text capabilities for Tamil audio and is distributed on Hugging Face for use with the Transformers library.

This is a fine-tuned version of the Whisper large-v3-turbo model optimized for German automatic speech recognition. It provides high-accuracy transcription for German-language audio. The model is available on Hugging Face and can be used with the Transformers library for local or hosted inference.

Whisper Tiny is an open-source automatic speech recognition (ASR) model developed by OpenAI. It enables developers to transcribe audio files into text across multiple languages, optimized for lightweight and efficient inference. The model is suitable for integration into speech-to-text pipelines and offline transcription tools.