AI actress in 2026: Tilly Norwood, Hollywood's pushback and the rules on AI‑generated actors

Short answer: an AI actress is a performer who exists only as software. Her face, voice and movements are generated by AI models rather than filmed from a real person. The best-known example is Tilly Norwood, an AI-generated actress made by the London studio Particle6, which announced in July 2026 that she will lead a feature film called Misaligned.

You won't find a ranked list of "the best AI actresses" here, because that list doesn't really exist. There are a handful of synthetic characters and many tools that can produce them. So we've written this as an explainer: who Tilly Norwood is, how AI-generated actors are made, what Hollywood and SAG-AFTRA have done about them, and the consent and likeness rules you need to know before you build a synthetic performer of your own. We compiled it from union statements, contract summaries, law-firm analyses and news reports, as of late October 2026. Nobody at the site tested these tools hands-on for this piece.

Key takeaways

• Tilly Norwood is a character, not a person. She was created by Particle6 and its AI talent studio Xicoia, both run by actor and producer Eline van der Velden.
• Hollywood's main union is openly opposed. SAG-AFTRA said Tilly "is not an actor, it's a character generated by a computer program" and that its contracts require notice and bargaining before a synthetic performer is used.
• The 2026 contract added limits. Producers say they won't use synthetic performers in human roles unless they add "significant additional value" to the production.
• Likeness law is moving fast. New York now requires ads to disclose synthetic performers, California restricts vague digital replica contracts, the federal NO FAKES Act cleared a Senate committee in June 2026, and Denmark is legislating to protect faces and voices.
• Consent is the line that matters. A fully invented face is one thing. Copying a real person's face or voice without permission is where legal and ethical trouble starts.

What is an AI actress?

An AI actress, or AI-generated actor more broadly, is a digital character built to look and behave like a human performer. There is no actor on set. Instead, image models design the face, video models animate it, and voice models give it speech. A human team still writes the scripts, directs the scenes and picks the best outputs.

It helps to separate three things that often get lumped together:

The labels matter because the rules differ. TheWrap reports that the 2026 SAG-AFTRA deal defines synthetic performers as "performers created completely from generative AI that are not replicas of a specific performer." Digital replicas of real people sit under separate, stricter consent rules.

Who is Tilly Norwood, the AI-generated actress?

Tilly Norwood made her debut at the Zurich Summit in September 2025. Her creator, Eline van der Velden, said there that she wanted Tilly "to be the next Scarlett Johansson or Natalie Portman," according to The A.V. Club. Talk of Tilly getting a talent agency set off a wave of anger from actors.

Van der Velden pushed back on social media. She wrote that Tilly "is not a replacement for a human being, but a creative work—a piece of art," and described AI as "a new tool, a new paintbrush." She also said: "I'm an actor myself, and nothing—certainly not an AI character—can take away the craft or joy of human performance."

Since then, Particle6 has kept the project going and leaned into the controversy:

• March 2026: Tilly appeared in a music video, "Take the Lead," that answered the backlash. Variety reports that the song was generated with Suno, that van der Velden performed Tilly's role, and that 18 people worked on the video. Van der Velden said: "Tilly is, and has always been, a vehicle to test the creative capabilities and boundaries of AI."
• July 2026: Particle6 announced Misaligned, a comedy-drama set in the "Tillyverse," a digital world in the cloud. The film is in early development. Van der Velden said it "will absolutely be funny, chaotic and self-aware -- very Tilly."

Note what's missing from the record. In the reports we read, there's no confirmed talent agency deal, no release date for Misaligned and no named studio distributor. Treat claims beyond that with caution.

Verdict: Tilly is less a working actress than a long-running publicity test of what AI performers can do and how audiences react.

How AI-generated actors are made

Studios don't publish full recipes, but the general pipeline is well understood. Particle6, for example, told Variety it used "a suite of AI tools" plus its own proprietary process. In practice, most synthetic actors are built in five stages:

1. Design the face. An image generator produces hundreds of candidate faces. The team picks one and locks it in so it stays consistent across shots.
2. Build a reference set. The chosen look is rendered from many angles, outfits and lighting setups. This becomes the character's "model sheet."
3. Animate it. Image-to-video and text-to-video models turn stills into moving scenes. Some teams drive the character with a human performance, so a real actor's movement and timing sit underneath the AI face.
4. Give it a voice. A synthetic voice is generated or designed, then synced to the face. Music, as with Tilly's song, can come from AI music tools.
5. Edit and direct. Humans choose takes, fix glitches and cut the final piece. As van der Velden put it, "great AI content isn't instant — it always takes good ideas, taste, direction, judgement and time."

If you want the technical side of the video step, see our explainer on how AI creates videos.

Hollywood's reaction: SAG-AFTRA and the AI actress debate

The strongest response came from SAG-AFTRA, the main US union for film and TV performers. Its statement on Tilly, quoted by The A.V. Club, was blunt:

"To be clear, 'Tilly Norwood' is not an actor, it's a character generated by a computer program that was trained on the work of countless professional performers—without permission or compensation."

The union added that "it doesn't solve any 'problem'—it creates the problem of using stolen performances to put actors out of work." It also reminded producers that its contracts "require notice and bargaining whenever a synthetic performer is going to be used."

That last point is the practical one. SAG-AFTRA isn't relying on outrage alone. It's relying on contracts.

What the 2023 contract already said

The deal that ended the 2023 strike set the first ground rules. According to an AFP report carried by eNCA, studios must get an actor's consent each time a digital replica is used, and must give a "reasonably specific description" of the intended use. Replicas can't be used to avoid hiring background actors. For synthetic performers built from features of several real actors, studios need permission from each actor whose features are used and must notify the union. Negotiator Duncan Crabtree-Ireland summed it up: "If you're using Brad Pitt's smile and Jennifer Aniston's eyes, both would have a right of consent."

What the 2026 contract added

SAG-AFTRA members ratified a new TV/theatrical agreement on June 4, 2026, with 91.42% voting in favour. It runs from July 1, 2026 to June 30, 2030. Based on TheWrap's summary of the deal, the AI changes include:

• A "human first" principle. The union says the deal rests on a principle "strongly favoring human performances."
• A higher bar for synthetics. Producers state they don't intend to use a synthetic performer in a human role unless it adds "significant additional value" to the production.
• No replicas as strike-breakers. Producers can't use a performer's digital replica to replace them during a strike.
• A reason to scan. Producers need an "articulable business reason" before scanning someone to create a replica.
• Extra protection for minors. The contract adds negotiated protections for digital replicas of child performers.

SAG-AFTRA president Sean Astin said the deal "strengthens protections around artificial intelligence and digital identity." Los Angeles Magazine notes that an arbitration provision reportedly limits AI use to a handful of edge cases, language the union had argued over. How the "significant additional value" test works in practice is the thing to watch.

Verdict: if you plan to work with a union production, a synthetic actor is not a loophole. It's a negotiation.

Consent and likeness rights: the laws that apply

Union contracts only bind union productions. For everyone else, including independent creators, advertisers and brands, the law is what matters. Here's where things stand as of October 2026. This is a summary, not legal advice.

The NO FAKES Act

Holland & Knight reports that the Senate Judiciary Committee advanced the NO FAKES Act by voice vote on June 18, 2026. The bill would give every person, famous or not, a licensable federal right over digital replicas of their voice and visual likeness. People and companies that make or distribute unauthorised replicas could be liable, and so could platforms that host them knowing they're unauthorised. It includes a takedown process modelled on the DMCA and carve-outs for news, parody and criticism. It still needs a full Senate vote, and the House version hasn't been taken up.

New York's synthetic performer rule

This is the law most directly aimed at AI actresses. Cooley explains that Governor Kathy Hochul signed it on December 11, 2025, and it took effect on June 9, 2026. Advertisers who know an ad contains a synthetic performer must "conspicuously disclose" it. A synthetic performer here is a digitally created asset meant to look like a human performance that isn't recognisable as a real person. Penalties are $1,000 for a first violation and $5,000 for each one after that.

California's replica laws

California moved early. Manatt's summary says AB 2602 makes a contract clause unenforceable when it allows an AI replica to replace work a performer could have done in person, the use isn't reasonably specific, and the performer had no lawyer or union representation. AB 1836 requires consent before making a digital replica of a deceased performer, with damages of at least $10,000. Both took effect on January 1, 2025.

Denmark's face-and-voice bill

Europe has its own approach. Euronews reports that Denmark's bill would amend copyright law to ban sharing realistic deepfakes of people's faces and voices without consent, covering ordinary people as well as artists. Satire and parody are exempt. Culture minister Zenia Stampe said: "One of the main goals of the bill is to get the tech giants to take down manipulated content quickly." For the wider picture, see our guide to AI regulation in 2026.

Tools people use to create synthetic actors

There's no single "AI actress app." Synthetic performers are stitched together from general-purpose tools. If you're exploring this for marketing, training videos or a short film, these are the categories you'll work with:

The more responsible vendors publish their rules. Synthesia, for example, lists "consent safeguards for creating custom avatars and voices" and says it moderates "content at the point of creation, not clean up after distribution." It also runs a programme it describes as "fair compensation for the actors behind stock avatars." That's the model to look for: real actors who agreed to be avatars and get paid for it.

Choose an avatar platform if you need a presenter who speaks to camera. Choose a video generator if you need a character inside a story. Use both if you're building a recurring character like Tilly.

How to use AI actors responsibly

If you're a creator or a brand thinking about an AI-generated actress, these rules will keep you on the right side of the law and your audience:

1. Never copy a real person without written consent. That includes "lookalikes" designed to evoke a specific celebrity. Making deceptive or intimate images of real people without consent causes real harm and is illegal in many places.
2. Disclose it. In New York ads, disclosure is now the law. Everywhere else, it's good practice and it protects your credibility.
3. Read the contract. If you hire a real actor to drive or voice a character, spell out exactly how their likeness will be used. Vague, perpetual rights are what California's AB 2602 targets.
4. Check your production's union status. On a SAG-AFTRA signatory project, using a synthetic performer triggers notice and bargaining.
5. Keep humans in the credits. Writers, directors, editors and any performers underneath the AI deserve to be named. Tilly's own music video had an 18-person team.

Working with real faces rather than invented ones? Our guide to AI face swap video tools covers the consent rules for face-based editing in more detail.

Final verdict

AI actresses are real, but they are still mostly experiments and marketing stunts. Tilly Norwood has generated far more headlines than screen time, and her first feature is still in development. The bigger story is the rulebook forming around her: union contracts that favour human performers, a New York disclosure law, California's replica limits and a federal bill waiting for a Senate vote.

If you only remember one thing, remember this. An invented face is a creative choice. A copied face without consent is a legal problem. Build on the first, stay well away from the second, and tell your audience what they're watching.

FAQ

Is Tilly Norwood a real person?

No. Tilly Norwood is an AI-generated character created by Particle6 and its AI talent studio Xicoia, run by Eline van der Velden. Van der Velden herself performed Tilly's role in the 2026 music video "Take the Lead." SAG-AFTRA calls Tilly "a character generated by a computer program," not an actor.

What film is AI actress Tilly in?

Particle6 announced in July 2026 that Tilly will star in Misaligned, a comedy-drama set in a digital world called the "Tillyverse." The film was in early development when announced. No release date or distributor had been confirmed in the reports we reviewed as of late October 2026.

Can AI-generated actors legally replace human actors?

Outside union productions, no law bans synthetic performers outright. But SAG-AFTRA's 2026 contract requires notice and bargaining, and producers say they won't use synthetics in human roles unless they add "significant additional value." Ads in New York must disclose them. Copying a real actor's likeness needs consent in many places.

How are AI videos with AI actors made?

Most are built in stages. An image model designs a consistent face, a video model animates it into scenes, a voice model adds speech, and human editors pick and polish the best takes. Some teams also drive the character with a real performer's movement. Our guide to AI video generators covers the main tools.

Which AI is best for making videos with a synthetic presenter?

For a presenter talking to camera, AI avatar platforms such as HeyGen or Synthesia are built for that job. For characters inside a story, general AI video generators fit better. Whichever you choose, pick a vendor that verifies consent before creating an avatar or voice of a real person.

Do I need consent to make an AI version of a real person?

In practice, yes. California requires estate consent for replicas of deceased performers, Denmark is moving to ban sharing realistic deepfakes without consent, and the proposed federal NO FAKES Act would create a right to sue over unauthorised replicas. Even where the law is unclear, using someone's likeness without permission is risky and harmful.