Credits

Lip Service is made from other people's generous work. Thank you.

The voices

The sixteen voices learned to speak from LibriTTS-R (Yuma Koizumi, Heiga Zen, Shigeki Karita, Yifan Ding, Kohei Yatabe, Nobuyuki Morioka, Michiel Bacchiani, Yu Zhang, Wei Han, and Ankur Bapna): volunteers reading public-domain books aloud for LibriVox, restored and published as a corpus under CC BY 4.0. Each voice is tuned to one reader and has an invented name.

The singing

How a held note moves, and the breaths singers took between notes, come from VocalSet (Julia Wilkins, Prem Seetharaman, Alison Wahl, and Bryan Pardo): recordings of trained singers, published under CC BY 4.0.

The words

Pronunciations come from the CMU Pronouncing Dictionary, from Carnegie Mellon University.

The rest

The fast sound uses the weights of Vocos (Hubert Siuzdak; MIT license). The speech model started from LJ Speech (Keith Ito and Linda Johnson), a public-domain recording. BERT (Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova; Apache-2.0) helped teach the model that places pauses. How it leans on an emphasized word was measured with the Crowdsourced LibriTTS Speech Prominence Annotations (Max Morrison; CC BY 4.0). The name is set in Outfit (The Outfit Project Authors; SIL Open Font License).

Every source, its authors and its license are listed in full in NOTICE.