The best speech recognition you can run at home is absolutely tiny, and it runs on a ten-year-old laptop
Parakeet is a tiny speech recognition model that transcribes on a ten-year-old laptop with no subscription
Parakeet is a compact speech recognition model that can be run on a wide range of devices, from Raspberry Pis to ten-year-old laptops, without the need for a subscription or powerful hardware. Developed by Nvidia, this 600 million parameter automatic speech recognition model delivers accurate transcription even in the presence of accents.
The model is lightweight, consuming only 700MB of storage and 2GB of RAM, and can be easily deployed using Docker or a Mac-friendly desktop app called MacParakeet. MacParakeet offers a user-friendly interface, complete with system audio and microphone input support, speaker identification, and the ability to transcribe web-based audio.
Transcripts can be exported in Markdown format, ensuring complete privacy as all data remains on the device. Parakeet's performance has been impressive in various scenarios, such as transcribing outlines for articles while walking on a treadmill, replying to emails, and working with personal notes. Compared to commercial solutions, Parakeet offers a resource-efficient alternative that maintains the quality needed for transcription purposes.
Written by urgent.news from XDA Developers's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
