Whisper Base

Whisper Base is the 74-million-parameter ASR model from OpenAI, providing robust, open-source multilingual speech-to-text transcription and translation.

This is the Whisper 'Base' model: a 74-million-parameter version of OpenAI’s Automatic Speech Recognition (ASR) system. It utilizes a Transformer-based encoder-decoder architecture, trained on a massive 680,000 hours of diverse, multilingual, and multitask supervised data. The extensive training set ensures superior robustness against common issues like background noise, accents, and technical jargon. The model efficiently handles core tasks: multilingual speech recognition, language identification, and speech translation from various languages directly into English.

https://github.com/openai/whisper