A “Latin American Singing Voice” Created with Artificial Intelligence

“Sometimes with sorrow, sometimes with hope” was presented, a song created from artificial intelligence with a “LATAM Singing Voice”, which brings together the voices of different popular singers from the region. The project, which is described in detail in the USO Models chapter of the IA + Music in Latin America report, was carried out collaboratively by Amplify artists (Berenice Llorens and Pol Lain) and the Larrosa and Futurx teams. From its conception, it sought to transcend the cloning of a conventional voice. An invention was proposed with the potential to combine the power and artistry of many of our singer-songwriters into a single speech,” they said in a press release. Interviews and songs by Violeta Parra, Mercedes Sosa, Gilda, Selena Quintanilla, Chavela Vargas, Tita Merello and copleras from the Argentinian northwest were the dataset to create a unique and completely new expression that contains the essence of each and every one of the artists. In addition, to carry out the project, a lyric with copla meter was created using ChatGPT4 and analyzing the perceptions about artificial intelligence of the surveys carried out for the report AI + Music in LATAM.  The rhythmic, timbral and melodic base was the result of an exploration of sound textures from the audios generated by Ktonal’s AI. 

Original source in Spanish