The Infinite Star
In 1994, Natalie Cole performed a duet with her father Nat King Cole—who had died in 1965. Using studio technology, his voice was extracted from old recordings and merged with hers. The result, "Unforgettable," won multiple Grammy Awards. Audiences wept.
That technological feat required months of studio work by skilled engineers. Today, you can clone someone's voice in minutes with a consumer AI tool. Feed it samples, and it speaks whatever text you provide. The voice of anyone—living or dead, consenting or not—can say anything.
This is the promise and peril of synthetic celebrities: AI that can simulate any person, with their appearance, voice, and mannerisms. Entertainment without limits. Performance by the dead. New content from artists who've stopped creating. Digital humans who never existed but feel completely real.
And also: fraud, deception, exploitation. Putting words in mouths without consent. Profiting from identities without permission. A world where you can't trust that any video is real.
This chapter explores the emerging reality of synthetic celebrities: what's already possible, who controls it, and what happens when anyone can simulate anyone.
2026 Snapshot — Synthetic Media Capabilities
Voice Cloning
Consumer tools: ElevenLabs, Resemble.ai, Descript Overdub, and others. Clone a voice with minutes of audio.
Quality: Near-indistinguishable from real for short clips. Longer content may have artifacts.
Uses: Dubbing, audiobooks (authorized), voice preservation, podcasting, scams (unauthorized).
Detection: Difficult and getting harder. Arms race with generation.
Face Generation and Animation
Image generation: Midjourney, DALL-E, Stable Diffusion can generate photorealistic faces of non-existent people.
Deepfakes: AI can map one person's face onto another's body in video. Quality improving; becoming accessible.
Character animation: AI can animate faces from audio alone. Lip sync matching speech.
Digital humans: Full synthetic characters for games, virtual assistants, marketing.
Full Synthetic Video
Video generation: Runway, Pika, Sora (OpenAI) generate video from text prompts. Quality improving rapidly.
Character consistency: Maintaining consistent characters across video remains challenging but advancing.
Real-time generation: Not yet possible for high quality; coming.
Text and Behavior
Writing style mimicry: AI can learn someone's writing patterns and generate text in their style.
Conversational clones: AI trained on someone's communications can conduct conversations "as" them.
Behavioral prediction: Given enough data, AI can predict how someone would respond to situations.
Notable Players
Technology Providers
Voice synthesis:
- ElevenLabs: Leading voice cloning
- Resemble.ai: Voice AI for media
- Descript: Overdub feature for podcasters
- WellSaid Labs: Enterprise voice
Visual synthesis:
- Midjourney, OpenAI, Stability AI: Image generation
- Synthesia: AI video generation with digital avatars
- Soul Machines: Digital human creation
- Hour One: Video generation for enterprise
- Runway, Pika, Luma: Video generation
Platforms:
- Character.ai: Conversational AI characters
- Replika: AI companion app
- Paradot, Kindroid: Similar companion apps
Entertainment Industry
Studios and labels:
- Major studios exploring AI for de-aging, resurrection, background characters
- Music labels negotiating AI rights with artists
- Estates of deceased artists licensing likeness
Talent representation:
- CAA, WME building AI/digital licensing capabilities
- SAG-AFTRA negotiated AI protections in 2023 strike
- Athlete and influencer agencies adding digital rights
Rights and Enforcement
Rights holders:
- Artist estates controlling posthumous use
- Athletes protecting NIL (name, image, likeness)
- Celebrities pursuing unauthorized uses
Legal frameworks:
- Right of publicity (varies by state/country)
- Copyright (complex application to AI)
- Emerging synthetic media laws
Current Applications
Authorized Uses
Dubbing and localization: AI voice cloning allows actors to "speak" foreign languages without dubbing actors.
Audiobooks: Authors can clone their voices to narrate books without studio time.
Legacy content: Estates authorize AI to create new content from deceased artists.
De-aging and enhancement: Films use AI to de-age actors or enhance performances.
Virtual influencers: Created characters (Lil Miquela, etc.) with millions of followers.
Unauthorized Uses
Fraud: Voice cloning for family emergency scams. CEO impersonation for wire fraud.
Political manipulation: Deepfakes of politicians saying things they didn't.
Non-consensual content: Synthetic pornography using celebrities' and ordinary people's faces.
Identity theft: Using someone's synthetic likeness for endorsements, scams, impersonation.
Gray Areas
Parody and commentary: Synthetic media satirizing public figures. Fair use? Defamation?
Fan content: Fans creating content with artists' voices/faces. Tribute or theft?
Historical recreation: Putting words in historical figures' mouths. Education or misrepresentation?
Deceased persons: Who controls the identity of someone who can't consent?
The Economics of Synthetic Identity
The Value of Identity
Celebrity identity is valuable: Endorsements, appearances, licensing worth billions annually.
Scarcity creates value: A celebrity can only be in one place, make one endorsement. Attention is finite.
Synthetic changes this: AI can create unlimited content. The celebrity doesn't need to be present.
New Revenue Streams
Infinite appearances: License your likeness for ads, movies, games—without being there.
Posthumous earnings: Estates can continue generating content and revenue indefinitely.
Personalized content: Fans could receive personalized messages "from" celebrities at scale.
Multiple "versions": Different markets could have different "versions" of a celebrity.
Economic Disruption
Background actors eliminated: AI can generate crowd scenes, background characters.
Voice actors challenged: AI voices for games, audiobooks, ads.
Human performance devalued?: If AI can perform, what's the premium for human?
Or enhanced: Authentic human performance may become more valuable when AI is ubiquitous.
Rights and Control
Current Legal Framework
Right of publicity: In most US states, individuals control commercial use of their identity. But:
- Varies by state
- Doesn't clearly cover AI-generated content
- Doesn't survive death in all jurisdictions
- Exceptions for news, commentary, parody
Copyright: Doesn't clearly protect appearance or voice. May protect specific performances.
Contract: Celebrities can negotiate AI rights in contracts. Industry standards emerging.
Emerging Protections
State laws: Tennessee ELVIS Act (2024) specifically addresses AI voice cloning. California, New York, others considering.
Federal proposals: NO FAKES Act and others proposed in Congress. Not yet passed.
Industry agreements: SAG-AFTRA 2023 contract requires consent for AI likeness use.
Platform policies: Some platforms restrict synthetic media; enforcement varies.
Key Questions
Who owns a likeness after death?: Estates? Descendants? The public domain?
What uses require consent?: Commercial? All? What about parody and commentary?
How is consent verified?: When AI content claims authorization, how is it verified?
What about ordinary people?: Celebrity protections don't cover everyone. Non-consensual deepfakes proliferate.
Synthetic Identity Markets
What's Emerging
Identity licensing platforms: Companies that manage AI rights for celebrities and estates.
Digital twin creation: Services that create authorized AI versions of people.
Consent registries: Systems to verify whether someone has authorized AI use of their identity.
Enforcement services: Companies that detect and pursue unauthorized synthetic media.
Market Structure Possibilities
Centralized licensing: A few large platforms control most celebrity AI rights. Like music licensing.
Decentralized: Many platforms, registries, enforcement entities. Fragmented.
Individual control: Each person manages their own AI rights. Complex but privacy-preserving.
Corporate ownership: Studios, labels own AI rights to talent. Extends employer control.
Economic Scale
Current: Small but growing. Mostly in entertainment and advertising.
Potential: If synthetic celebrity becomes routine, market could be billions annually. Every ad could feature any celebrity.
The Path Forward
Near-Term Likely (2026-2032)
Quality improves: Synthetic video indistinguishable from real for short clips. Voice cloning perfect.
Authorized use expands: More celebrities license AI rights. Studios use for dubbing, de-aging, background.
Unauthorized use proliferates: Deepfake scams, non-consensual content increase. Platforms struggle to enforce.
Legal frameworks develop: More states pass synthetic media laws. Industry standards emerge. Federal law debated.
Detection arms race continues: Better detection, better generation. No clear winner.
Plausible (2032-2040)
Synthetic celebrities are normal: AI-generated content featuring celebrities is routine in advertising, games, media.
Licensing infrastructure matures: Clear systems for authorizing, tracking, compensating AI use of identity.
Deceased performers return: New movies starring long-dead actors. New albums from deceased musicians. Ethical debates ongoing.
Consent verification standard: Systems exist to verify whether synthetic media is authorized.
Ordinary people get protection: Laws and tools extend to non-celebrities.
Wild Trajectory (2040+)
Anyone can appear in anything: Perfect synthetic generation means any content can feature anyone.
Identity markets are huge: Trading in identity rights like intellectual property. Speculation on celebrity value.
Human celebrity is devalued: Why hire a human when AI is indistinguishable and cheaper?
Or: Human authenticity premium emerges. Live performance, verified human creation become more valuable.
Or: Consent becomes meaningless. Everyone appears in everything. Identity as currently understood dissolves.
Risks and Guardrails
Non-Consensual Use
Risk: AI used to create content without permission. Harassment, exploitation, fraud.
Guardrails: Legal protections (right of publicity expansion); platform enforcement; detection tools; consent verification; criminal penalties for malicious use.
Fraud and Deception
Risk: Fake videos of public figures; fake endorsements; fake evidence; financial fraud.
Guardrails: Authentication standards (C2PA); disclosure requirements; fraud prosecution; public education.
Power Concentration
Risk: Studios/labels control AI rights; talent has no bargaining power; exploitation.
Guardrails: Union protections; mandatory consent; term limits on AI licensing; regulatory oversight.
Cultural Manipulation
Risk: Synthetic media shapes culture without disclosure. Manufactured movements, fake popularity.
Guardrails: Disclosure requirements for synthetic media; transparency about AI-generated content; media literacy.
Identity Erosion
Risk: When anyone can be simulated, identity itself loses meaning. Trust in all media collapses.
Guardrails: Content authenticity infrastructure; verified human attestation; social norms against unconsented use; acceptance of uncertainty.
The Deeper Questions
Who Owns Identity?
Identity feels like something each person owns—face, voice, manner. But is that actually the case?
The law has treated identity as property-like (right of publicity) but also as connected to dignity and personhood. AI forces the question: is identity something to be licensed, traded, and exploited? Or something sacred that shouldn't be commodified?
What Survives Death?
When someone dies, what happens to their identity? Their estate may control physical property, but what about the ability to create new performances, new statements, new appearances?
Currently, this varies by jurisdiction and contract. But the underlying question is philosophical: do the dead have interests? Can they be harmed by posthumous use? Who should decide?
Is Authenticity Valuable?
If AI can create indistinguishable content, does it matter whether a human created it?
Some argue no—quality is quality. Others argue authenticity has intrinsic value—knowing a human made something changes its meaning.
The market will reveal preferences. But preferences can be shaped. What people value depends partly on what they're taught to value.
Where Is the Line?
Should an individual be able to create a private video featuring a favorite celebrity for personal enjoyment? What about sharing it? What about selling it? What about a video that depicts violence or intimacy without consent?
Different people draw lines differently. The technology doesn't care—it can create anything. Society must decide what's permitted, what's prohibited, and how to enforce the distinction.
Conclusion
Nat King Cole's duet with his daughter was a marvel—the technology that brought a dead man's voice to life. It was also a choice: his estate authorized it, his daughter performed it, the result was meant to honor his legacy.
The coming world of synthetic celebrities will force many such choices, multiplied by millions. Every public figure—and eventually every person—will face questions about how their likeness can be used, by whom, for what purposes.
The technology itself is neither good nor evil. It enables tribute and exploitation, creativity and fraud, connection and deception. What matters is how it is governed: what rights are granted, what consent is required, what accountability is enforced.
The stakes are high. Identity is fundamental to how people understand themselves and each other. A world where anyone can be made to say or do anything is a world where nothing can be trusted. A world with clear rules about synthetic identity might preserve trust while enabling creativity.
Society is not yet in either world. It remains in between, making choices that will determine which it becomes.
Endnotes — Chapter 39
- "Unforgettable" duet by Natalie Cole with Nat King Cole released 1991; won Grammy for Record of the Year and Album of the Year.
- ElevenLabs voice cloning requires approximately 1 minute of audio for basic cloning; higher quality with more samples.
- Deepfake technology began with academic research (2014-2017); became publicly accessible around 2018; quality has improved dramatically since.
- Lil Miquela: virtual influencer created by Brud (now Dapper Labs); 3+ million Instagram followers; appeared in ads for major brands.
- SAG-AFTRA 2023 contract negotiations included AI provisions requiring consent for digital likenesses; strike lasted 118 days.
- Tennessee ELVIS Act (Ensuring Likeness Voice and Image Security Act) passed 2024; specifically addresses AI voice cloning without consent.
- NO FAKES Act (Nurturing Originals, Fostering Art, and Keeping Entertainment Safe) proposed in US Congress; addresses AI-generated content using likeness.
- Right of publicity varies significantly by state; some recognize perpetual rights after death (e.g., California), others limited or none.
- Voice cloning scams documented by FTC and media; "grandparent scams" using cloned family voices to request emergency money.
- Digital resurrection in entertainment includes examples like Peter Cushing in "Rogue One" (2016), young Luke Skywalker in "The Mandalorian," and various music-related projects.