Play.ht published a fictional podcast interview between Joe Rogan and Steve Jobs, using synthesized voices for both speakers. Ars Technica reports this.
\n\n\n
The podcast was created using voice cloning technology. The developers used deep learning to reproduce the distinctive features of speech with high accuracy.
\n\n\n
Engineers trained the model on existing voice samples. According to the developers, Rogan’s voice is easier to clone, as there is a sufficient amount of material publicly available suitable for training.
\n\n\n
Jobs’s voice came out less realistic and more robotic. At times it resembles the former Apple CEO’s appearances and All Things Digital interviews from the late 2000s.
\n\n\n
The interview text was also generated by AI. The developers may have used a large language model such as GPT-3.
\n\n\n
“For example, the episode with Steve Jobs was trained on his biography and all the records we could find on the Internet, so the AI could accurately bring him back to life,” Play.ht said.
\n\n\n
The 19-minute interview does not contain meaningful information. After some time, portions of the fictional podcast begin to sound like conceptual blends of Jobs’s general themes, including aesthetics, revolutionary products and competitors.
\n\n\n
“This is a problem I have always faced at Microsoft. In many respects they are smart people and have done a good job, but they have never had taste. They have never had any aesthetic sense,” says the fake Jobs.
\n\n\n
The full interview is available on the доступно on the Transistor podcast site.
\n\n\n
In September 2022, the Ukrainian startup Respeecher granted permission to clone the voice of actor James Earl Jones for voicing his Darth Vader character in future Star Wars projects.
\n\n\n
In August 2021, the company Sonantic used artificial intelligence to recreate the voice of Val Kilmer, who had lost it due to throat cancer.
\n\n\n
Subscribe to ForkLog news on Telegram: ForkLog AI — all the news from the AI world!