Wake Words

Wake Words Real recordings of Jewish and Israeli first-name wake words, for training local wake word models (openWakeWord, microWakeWord) to use with Home Assistant Assist and on the desktop, in place of "Hey Siri" or "OK Google". Wake word wake_word Status Hey Yossi hey_yossi Main word: about 30% of recordings and the most takes per speaker Hey Mendel hey_mendel Secondary Hey Herman hey_herman Secondary Hey Doug hey_doug Secondary Hey Reuben hey_reuben… See the full description on the dataset page: https://huggingface.co/datasets/danielrosehill/Wake-Words.

About This Project

Wake-Words is a specialized audio dataset containing recordings of specific first names intended for training local wake word models. It supports the development of custom voice triggers for platforms like Home Assistant Assist and desktop applications, serving as a private, local alternative to commercial voice assistants.

The dataset includes a variety of recordings for names such as Yossi, Mendel, Herman, Doug, and Reuben, structured to support machine learning frameworks like openWakeWord and microWakeWord. By providing categorized audio samples in both English and Hebrew, the project enables developers to implement personalized voice activation triggers that run entirely offline.

This resource is designed for users and developers working on home automation, privacy-focused voice interfaces, and custom keyword spotting tasks. It offers a structured foundation for those seeking to move away from proprietary cloud-based wake word recognition toward self-hosted, localized solutions.

Project Details

Tags

task_categories:audio-classificationlanguage:enlanguage:helicense:cc-by-4.0size_categories:n<1kformat:parquetmodality:audiomodality:textlibrary:datasetslibrary:pandaslibrary:polarslibrary:mlcroissantregion:uswake-wordkeyword-spottinghotwordopenwakewordmicrowakewordhome-assistant

Explore More Projects