AI and audio
Last updated 16 November 2026
Every song, every sung line and every spoken word on Lyriday is generated with AI. We think you should know that before you press play.
How a song is made
We write the lyrics and choose the words for each lesson. ElevenLabs music models turn the lyrics and a style (for example bossa nova or city pop) into a sung song. ElevenLabs speech models say each word slowly. ElevenLabs alignment tools time every word to the singing, which is how the lyrics light up as you listen.
Checked before you hear it
Each song is listened to by a speech recogniser. If one of the lesson words is not heard clearly, the song is made again. On this site, every lesson word in all eight songs was heard back.
No real singers imitated
The voices are generated. They do not copy any real artist, and we never ask the models to sound like one.
Mistakes
Generated singing can stretch or bend a word. If you hear a word that sounds wrong, write to hello@lyriday.com and we will fix that lesson.
Rights
The songs are made under a paid ElevenLabs plan that allows commercial use of the output. They are published here for listening on lyriday.com.