ElevenLabs makes ElevenMusic free for people who have lost their voices: 11,000+ people in the program

Greg Kellner spent his life teaching music: playing guitar, singing and communicating through gestures. In 2019, brain cancer took away his singing voice, forcing him to put teaching on hold. Seven years later, he joined the ElevenLabs Impact program and is now singing again. The ElevenMusic model recreated his voice from old recordings. On November 11, he took the stage at the Summit in New York and performed new songs using that voice.

It is a moving story, but its significance goes beyond emotion. Behind it is a product strategy: ElevenLabs is systematically turning voice cloning from a paid feature into an accessibility tool.

Two voices: from the archive and from scratch

The announcement presents two scenarios that are technically different.

The first is restoration. A person has recordings of their former voice, which are used to build a digital replica. That was Kellner's case: material from the past became a voice he can use to write and release new songs.

The second is design. Abdi Omar has never spoken because of cerebral palsy. There was nothing to clone, so he built a voice from scratch in Voice Designer. Omar is now a rapper: ElevenMusic helps him write and produce tracks. It is a compelling case for generating voices from descriptions, a capability often dismissed as a toy for video voiceovers.

How the program works

Until now, 1 Million Voices has focused on speech. According to the company, it has helped more than 11,000 people with ALS, motor neuron disease, head and neck cancer and other conditions speak again in a voice resembling their own. Access is free for life. The ambition is substantial: to restore a voice to one million people. This is a $1 billion commitment in kind, meaning the estimated value of the company's own products and services, rather than cash.

Now singing has been added alongside speech. People with permanent voice loss will receive free access to ElevenMusic. The launch is being carried out with Songwriters of North America (SONA): the association verifies eligibility and distributes access to its members and the wider songwriting community. More partners are promised later.

Two new partners provide other routes to access:

• NYU Langone Health. The program is available through Rusk Rehabilitation to patients with ALS, head and neck cancer, those recovering from a stroke and people with other conditions. Applications can be made through their rehabilitation team.
• Wounded Warrior Project. The veterans' organization will hold an educational workshop with ElevenLabs on how voice AI can help people with voice loss, blindness and low vision. The two organizations will then explore how voice technologies could connect veterans and their families with services.

Notice the approach: the company is not building its own clinical infrastructure. It is connecting with organizations already in contact with patients: hospitals, nonprofits and associations. Those partners also handle verification, the most awkward part of the process.

What to keep in mind

This is an official announcement, and it contains almost no measurable data. There is no information on how many recordings restoration requires, how closely the model matches the real voice or how many people have received access specifically to music. The 11,000 figure refers to speech, not singing. The "$1 billion in kind" is the company's own internal valuation. And the WWP collaboration currently amounts only to exploring possibilities, rather than a finished product.

There is also a broader issue: a voice is biometric data, so cloning should be based on the person's consent. The announcement says nothing about this. Distributing access through clinics and professional associations could be interpreted as a way to prevent misuse, but that is our interpretation, not a statement from the company.

Practical takeaways

Product managers and founders should take three things from this story.

First: combining "restore from an archive" with "create from scratch" covers almost all users. Some have recordings and others do not, so a product needs both paths.

Second: distribution channels make all the difference. A clinic, a songwriters' association and a veterans' organization already have people's trust and know how to verify eligibility. A startup could spend years building that trust itself.

Third: accessibility and business are compatible. Free access for a narrowly defined group demonstrates a model's capabilities better than any promotional video. If you are building a voice product, consider which communities of people with permanent voice or speech loss could become your first partners.