Skip to content

ElevenLabs Voice AI Gave a Composer with ALS His Singing Voice Back

ElevenLabs recreated composer Patrick Darling's pre-ALS singing voice on Eleven Music, letting him perform live in London three years after he lost it.

ElevenLabs Voice AI Gave a Composer With ALS His Singing Voi
Credit: ElevenLabs

Patrick Darling stopped being able to sing three years ago. On February 11, 2026, in front of nearly 1,000 people at the ElevenLabs Summit in London, he sang again, in his own voice. The 32-year-old composer, once the singer of the Irish folk group The Ceili House Band, was diagnosed with Motor Neuron Disease, the condition also known as ALS, at 29. What he performed that night was an original song, "Ghost of a Man I Never Met," carried by a recreation of the voice he had before the disease took it.

ElevenLabs built the vocal from recordings Darling made before he lost his voice, shaped on its Eleven Music platform by Richard Cave, a speech therapist on the company's partnerships team. The company says the aim was to recreate the timbre and character of his own singing rather than hand him a generic synthetic voice, so what the room heard was recognizably him. He performed alongside former bandmates Nick Cocking and Hari Ma. "Losing my ability to sing had a profound and devastating effect on me. It felt like I was losing an essential part of my identity," Darling said. "For the first time in years, I could hear myself sing again."

The performance doubles as a demonstration for the ElevenLabs Impact Program, which offers this kind of voice recreation free to people with permanent voice loss and to nonprofits in healthcare, education, and cultural work, with a stated goal of reaching one million voices. A staged summit performance is the polished version of that work, and the harder test is whether the same result holds for people who are not a professional musician with clean pre-diagnosis recordings and a company team building the model by hand. What Darling got back was narrow and personal, one song in one voice on one night. For someone who thought that was gone for good, that was the whole point.

Share this story

Julian Beaumont

Julian Beaumont covers artificial intelligence and large language models for techshooked, following the path from research paper to deployed feature. His standard is anti-hype: ask what a model actually does, what trained it, how it fails, and whether a benchmark measures what the announcement claims.