Livekit Wake Word Audio Dataset

An audio dataset for wake word detection, likely associated with the Livekit platform. The dataset was created by the author 'yepher' and was last updated in March 2026. Specific details on the number of samples, audio length, and recording conditions are not provided.

Use Cases

• Train a binary classifier to detect the 'Hey Livekit' wake word from background noise and other speech.
• Benchmark model latency and accuracy for real-time audio stream processing using short audio clips.
• Fine-tune a pre-trained speech model on a specific trigger phrase to improve false acceptance and rejection rates.
• Analyze acoustic features like Mel-frequency cepstral coefficients (MFCCs) from the audio samples for keyword spotting.

Strengths

• Dataset is associated with a specific, modern platform (Livekit) for real-time audio/video.
• Last update recorded in 2026, indicating recent maintenance.

Limitations

• No metadata on sample size, audio duration, or speaker diversity is available, limiting reproducibility.
• Unknown recording quality, environment, and potential bias from a single collection source.

Provenance

Source

Hugging Face user 'yepher'.

Collection Method

Likely custom recordings or synthetic generation for the 'Hey Livekit' phrase; method is unspecified.

Freshness

Last updated March 2026.