Quick answer
AI voice cloning lets you produce convincing vocals in any singer's voice from a few minutes of training audio. As of 2026 it is legal to use on yourself or with explicit consent. It is increasingly illegal to use on a real artist without permission: Tennessee's ELVIS Act (2024) protects voice as an explicit attribute, and similar state laws are spreading. Federal law has not caught up.
How voice cloning works
The model is trained on a target speaker's voice samples. Modern systems need 30 seconds to a few minutes of clean audio. The training produces a vocal "embedding," a numeric fingerprint of that voice. The model can then synthesize new speech or singing using the fingerprint, fed any text or melody.
The hard parts are: maintaining the voice's identity across pitch range (singing is harder than speaking), getting natural breathing and phrasing, and matching emotional tone. State-of-the-art systems in 2026 handle these well enough that an untrained ear can't reliably tell a clone from a real recording.
The famous cases
Drake and The Weeknd (Heart on My Sleeve, 2023)
The first AI-cloned song to hit major streaming charts. Anonymous creator. Pulled within days. Full timeline here.
Taylor Swift deepfakes (2024)
AI-generated explicit images of Taylor Swift went viral on X in January 2024. The platform was slow to remove them. Public outcry pushed the U.S. White House to call for federal AI deepfake legislation. Music industry voices used the moment to push for vocal-identity protections in the same package.
The "AI Frank Sinatra" sessions (2024 to 2026)
Several producers have released collections of AI-cloned Sinatra vocals over modern instrumentals. The Sinatra estate has been notably permissive about AI experimentation under licensed terms. Other estates (Prince, Whitney Houston) have been hostile. The pattern: the artist's heirs or rights-holders decide case by case.
Drake's own use of AI (2024 to 2026)
Multiple Drake releases since 2023 have been suspected of using AI vocal cleanup or stems. Drake has not commented directly. The pattern across major artists is: AI tools used by the artist's team are normalized. AI tools used to mimic the artist are fought.
Where it's legal, where it isn't
| Use case | Status in U.S. (2026) |
|---|---|
| Cloning your own voice for your own songs | Legal everywhere |
| Cloning a public domain voice (Sinatra etc.) without rights-holder consent | Legal but high reputational risk |
| Cloning a living artist for parody or commentary | Generally protected under fair use, contested in court |
| Cloning a living artist for commercial release | Likely illegal in Tennessee (ELVIS Act) and a growing list of states. Federally unclear. |
| Cloning a living artist for explicit deepfakes | Newly illegal under multiple state laws and the federal DEFIANCE Act (2024). |
Tennessee's ELVIS Act, explained
The Ensuring Likeness, Voice, and Image Security Act passed Tennessee in 2024. It's the strongest U.S. statute protecting an individual's voice as a property right. Key points:
- Voice is named explicitly as a protected attribute (most state right-of-publicity laws cover name and likeness only).
- Liability extends to anyone who knowingly distributes a work that uses the voice without consent.
- Applies to AI-generated voices, not just sampled or impersonated ones.
Tennessee passed the law because Nashville is the country music industry's center of gravity. Similar bills are pending or passing in California, New York, Illinois, and Texas. Expect a state-by-state patchwork through 2027, then likely federal legislation.
Boulevard's position on voice cloning
Boulevard is the AI alternative to Spotify. Our catalog is generated AI music from artists who don't exist. We do not clone real artists' voices, by policy and design. The technology to do it would be straightforward. We don't because (1) it's the wrong listener experience, (2) it's a much worse business, and (3) the legal landscape is moving fast enough that any clone catalog today will be a liability tomorrow.
Where this is going
Three predictions for the rest of the decade:
- Federal voice-rights law passes. Probably late 2027 or 2028. The state patchwork makes a federal statute inevitable.
- Licensed clones become a market. Artists in their later careers will license their voice to AI companies for posthumous use. Several are already negotiating these deals quietly.
- Detection arms race continues. Spotify, Apple Music, and YouTube are investing in AI-content detection. Some of it works. Most catches obvious clones. Clean clones still slip through.
Skip the Spotify subscription. Try the AI alternative.
Boulevard is the AI music app. Free to start. Listen instantly in your browser.