Gloria has cast for twenty-two years, and the productions she serves now routinely receive submissions that include AI-generated synthetic performers alongside the actors she has spent decades knowing. She still books real people. She is running out of reasons the producers find persuasive.
Gloria has cast for twenty-two years. She has sat across tables from performers at every stage of their careers — the twenty-three-year-old reading for their first network pilot, the sixty-year-old character actor whose face the audience knows before they know his name, the child actor whose mother holds her hand in the lobby while Gloria talks with them about whether the schedule is manageable. She has a particular gift for the part of the job that is not in any formal job description: knowing which actor brings something to a character that the director hasn't yet articulated they need. She has been told this many times. She has also been told, with increasing frequency, that she takes too long.
The acceleration she is accommodating is not only a scheduling issue. It is structural. The productions she works with — primarily mid-budget streaming content — are now receiving, alongside the actor submissions she compiles, a separate submission from AI-assisted casting tools: composite character options generated from text descriptions, sometimes with voice, sometimes with motion capture applied to synthetic faces, sometimes with an actor's previously consented likeness used as a base. The terminology varies. The effect is consistent: Gloria is submitting real people for roles that may be filled by non-people, and she is not always the one who decides which way the decision goes.
"I sent four actors last Tuesday for a guest role," she said. "All four were right. All four could have done the part. The director chose the AI composite because the voice was 'closer to what he was hearing.' The role was twelve lines."
The New System
Last year I wrote about the writers' room that learned to use AI — the story of two premium streaming productions, one that embraced AI writing tools and one that refused, and the question of what the craft of storytelling loses and gains in that choice. Gloria's story sits one remove from that question, in the casting room where the output of someone's creative decision — to use synthetic performers for this category of role — becomes her daily working reality.
The AI performer market has developed rapidly since the SAG-AFTRA AI consent provisions were incorporated into major production contracts in 2023. The provisions established, for the first time, a framework for the use of performers' digital likenesses: consent requirements, residual structures, use limitations. The provisions are meaningful and were hard-won. They cover members of the union who have likenesses to consent with. They do not cover the AI-generated composites that are built without any particular performer's likeness as a base — synthetic faces trained on datasets that may include photographs of working performers who never consented to that training, but who cannot be individually identified in the output.
The 18 percent figure for AI-generated background performance is an industry estimate, not a precise audit, and it is contested. SAG-AFTRA's estimate is based on production filings; industry observers note that many productions do not fully disclose their synthetic performer usage, and the actual figure may be higher. The direction of the trend is not contested.
What the Data Says
PwC's Global Entertainment and Media Outlook projects that the AI synthetic performer market will reach approximately $1.4 billion in production savings by 2028, driven by background performance, digital doubles for safety-sensitive stunt work, and localization — using licensed likenesses to dub or re-lip-sync foreign-language versions of content without requiring additional actor travel. The savings are real and the legitimate uses are real.
The Casting Directors' Association conducted a compensation survey in 2025 that found inflation-adjusted casting director fees had declined 11 percent since 2022, as productions became more likely to conduct initial AI-assisted casting passes that reduced the number of actors casting directors were asked to personally evaluate. The trend is most pronounced in background and smaller featured roles — exactly the roles that provide income continuity for the working actors and emerging talent that casting directors represent into the industry.
The distributional consequences of synthetic performer substitution are not evenly spread across the performer labor market. Stars with SAG-AFTRA contracts, established likenesses, and negotiating leverage are substantially protected. Background performers, newcomers who depend on featured background work for income while building credits, and character actors working in the mid-budget range are most directly affected by the substitution. Gloria casts in that range.
|
"I know five thousand actors. I have watched them develop. I have watched them become the thing they were always going to become. The composite doesn't become anything. It just performs." |
What the Worker Says Back
Gloria's professional identity is built around a particular epistemology: that knowing performers as human beings — knowing their range, their history, their capacity to surprise — produces better casting decisions than any system trained on visual and vocal data can produce without that knowledge. She believes this. She believes it not as a sentimental position but as a professional one, grounded in twenty-two years of evidence that the actor who seemed wrong for the role on paper was right for the role in the room, and that the room is not something a synthetic composite can be in.
She is aware that this position is hard to defend quantitatively. The director who chose the AI composite for the twelve-line guest role was not wrong that the voice was closer to what he wanted. He got what he was hearing. The question Gloria keeps returning to is what he didn't get — what the actor she submitted might have brought to those twelve lines that the composite, by design, was not capable of bringing. She does not know the answer. She knows the question matters.
"An actor brings their life," she said. "They bring what happened to them last Tuesday and what they're afraid of and what they're proud of and what they're trying to prove. It's in the performance, even in twelve lines. You can't train a composite on that. You can only get the surface."
Her concern is not only for the performers, though it is substantially for them. It is also for the work itself — for whether the accumulated replacement of small performances with synthetic approximations produces a discernible change in the texture of what audiences receive. She does not think anyone will be able to point to the moment it changed. She thinks it will change anyway.
The Contradictions
The streaming platform's approach to AI performer use is framed around cost efficiency and creative flexibility. The AI casting tools allow rapid iteration — a director can hear twenty voice options for a character before committing to a casting session, can visualize a character's look before scheduling screen tests. This is genuinely useful at the development stage of production. The problem is that the tools developed for the development stage have migrated into the production stage, because the cost differential is compelling and the oversight mechanisms are still catching up.
SAG-AFTRA has been aggressive in protecting its members, and the AI consent provisions represent a significant advance in a rapidly moving landscape. The provisions are also structurally limited by the boundary of union membership: they protect the performers who are already union members with established careers and likenesses. The performers at the beginning of their careers — who depend on featured background work, day player roles, and small guest roles for income and credits — are the ones most exposed to synthetic substitution and least protected by the existing framework.
Gloria casts those performers. She has given first credits to people who have gone on to careers she is proud of. She is increasingly uncertain how many of those credits she will be able to give to the next generation.
A Working Conclusion
Gloria had a session scheduled for that afternoon — a streaming production casting featured background performers for a market scene that would appear in three episodes. She estimated the role would involve two days of work per person, four to six people total. The production had originally requested AI composites for the scene. After she pointed out that the scene required reactive performance to dialogue from the lead cast — something that, as of current technology, requires human presence on set — they had agreed to use real performers.
"I'm getting better at the argument," she said. "I'm getting better at knowing which battles I can make the creative case for and which ones are just economics." She paused. "The economics are going in one direction. And the creative cases I'm able to make are getting smaller."
She picked up the session sheets. Fourteen actors' names, headshots, and reels. Twenty-two years of knowing how to read a person's potential in what they had done so far. She headed to the casting suite.
The composite database had four hundred thousand synthetic options available. None of them had an agent. None of them had been on the subway this morning, navigating the ordinary weight of their ordinary lives into the room where they would be asked to perform.
Figure 12. Stat callout card: '18% of background performance work now synthetic — with $320M in estimated production savings in 2026. Gloria still casts the humans. The question is for how long. (SAG-AFTRA / …
REFERENCES
1. SAG-AFTRA AI and Digital Replica Provisions: 2023 Television Agreement. SAG-AFTRA (2024).
https://www.sagaftra.org/contracts/tv-theatrical/artificial-intelligence
2. Global Entertainment and Media Outlook 2025–2029. PwC (2025).
https://www.pwc.com/gx/en/industries/tmt/media/outlook.html
3. Synthetic Performers and the Future of On-Screen Labor. Variety (2026).
4. Casting Directors in the AI Era: Compensation and Role Transformation. Casting Society of America (2025).
5. Digital Replicas, Consent, and Labor Rights in the Entertainment Industry. The Markup (2025).
6. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Kate Crawford (2021).
7. AI Likenesses and the Right of Publicity: Legal Frameworks for Performer Protection. Columbia Journal of Law and the Arts (2025).
Comments (0)
Join the conversation!