Datasetpublic

TigreGotico/synthetic-wakeword-hey_jarvis

synthetic-wakeword-hey_jarvis Synthetic wake-word audio for training and benchmarking OVOS wake-word plugins, covering the phrase "hey jarvis". Every clip is machine-generated text-to-speech. No human recording is included, and no natural voice is reproduced. Machine-generated audio carries no copyright of its own, so this dataset is published CC-BY-4.0 and is free to use, redistribute and build on, including for model training. Produced with support from the NGI0 Commons Fund.… See the full description on the dataset page: https://huggingface.co/datasets/TigreGotico/synthetic-wakeword-hey_jarvis.

sourceHugging Facecc-by-4.0updated 9d agoView on Hugging Face

0likes81downloads

Dataset Card

synthetic-wakeword-hey_jarvis

Synthetic wake-word audio for training and benchmarking OVOS wake-word plugins, covering the phrase "hey jarvis".

Every clip is machine-generated text-to-speech. No human recording is included, and no natural voice is reproduced. Machine-generated audio carries no copyright of its own, so this dataset is published CC-BY-4.0 and is free to use, redistribute and build on, including for model training.

Produced with support from the NGI0 Commons Fund.

Revision 2026-09-18: edge-tts renderings

Folder edge-tts-2026-09-18/: 1963 clips of the phrase "hey jarvis" made with edge-tts 7.2.8, one clip per combination of 40 English neural voices, 5 rates (-25 %, -12 %, 0, +12 %, +25 %), 5 pitches (-20 Hz to +20 Hz in 10 Hz steps) and 2 text forms (with and without an exclamation mark). Byte-identical renderings (37) were removed, so the count is below 2000. manifest.csv names the voice, rate, pitch, text and duration of every clip. 16 kHz mono PCM WAV, silence trimmed at 40 dB.

Seven voices were held out and appear in no clip, so an evaluation set can use them: en-US-AvaNeural, en-GB-MaisieNeural, en-IE-ConnorNeural, en-NZ-MollyNeural, en-SG-WayneNeural, en-KE-AsiliaNeural, en-IN-PrabhatNeural.

No voice conversion was applied to this revision. The licence is unchanged, CC BY 4.0: machine-generated audio, no human recording. The generator script is T-2909-gen_tts.py in the OpenVoiceOS agent workspace wiki (knowledge/wiki/audits/wakeword/). Produced with support from the NGI0 Commons Fund.