faster-whisper — быстрое распознавание речи в текст

faster-whisper is a fast implementation of Whisper speech recognition on the CTranslate2 engine, running noticeably faster and more memory-efficiently than the original. Reach for it when you need to transcribe speech to text locally at high speed: convert audio and video to text on CPU or GPU several times faster than plain Whisper, at the same accuracy — as a library to embed in your pipelines. Its focus is a fast, efficient ASR engine (speech → text) for developers, not a desktop app (buzz) or transcription with speaker diarization (whisperX, which often uses it under the hood): the value is efficient local transcription for embedding. It suits developers who need fast, resource-efficient speech transcription as a component of their apps and services.