The production survived weather, schedules, and three rewrites, and then you introduce AI into ADR. All at once, actor consent for AI voice is the tricky part, and the actor has good reasons to be wary.
So this piece walks through what actually goes into an AI voice cloning consent agreement, how to talk an actor through it without sounding like you’re hiding a trap in the fine print, and how to choose a provider that makes “yes” feel safe instead of turning every email into a mini‑grievance.
Key Takeaways
- Voice consent in film production is now a scope-and-term document covering what gets cloned, where it runs, how long, and how to exit.
- Many refusals come down to vague terms or a late start rather than the technology, so raising consent at casting and naming the boundaries settles the majority of objections early.
- What an AI voice agreement should include is now fairly settled: plain technology description, named intended use, compensation tied to scale, revocation rights, and a separate training-data clause.
Why Actors Say No (And What They're Really Afraid Of)
At the surface, “no” usually sounds like: “AI makes me uncomfortable,” or “my agent hates this stuff.” Underneath, the problem is almost always that the actor doesn't know what they're signing, how far it goes, or how they’ll stay in control once their voice is cloned.
The worry is a scenario they can picture: their voice turning up somewhere they never agreed to. A political ad, an awkward brand, a project that clashes with their values. Add in the feeling that AI can be duplicated endlessly, and “one session fee forever” suddenly looks like a bad bargain.
Most real fears cluster into three buckets:
- The voice keeps getting reused, but consent was one-time and vague.
- The voice is dropped into projects they’d never personally sign off on.
- The money doesn’t match the scale of use, especially when there’s global, long-tail distribution.
The Scarlett Johansson / OpenAI episode poured rocket fuel on all of this. An A‑list actor saying “I said no, and still ended up in an AI voice controversy” travels fast through agents’ chats. It’s a neat story that confirms every suspicion actors already had: if you don’t lock the rules down, someone else will decide what “reasonable use” means for you.
When an actor sees clear boundaries, what's being cloned, where it can run, how long, how to get out, they stop arguing with AI in general and start negotiating scope, projects, and rates. In other words, most “no”s are a very rational reaction to contracts that try to be “future-proof” by being vague.
What the Law and SAG-AFTRA Actually Require
If you’re using an actor’s voice as a digital replica, both SAG-AFTRA and California don’t treat that as a cute experiment. The rule of thumb: if your consent language could fit on a coffee mug (“for marketing, in all media, forever”), it’s probably not compliant anymore.
SAG-AFTRA's TV/Theatrical agreement treats digital replicas like a big deal, since it is one. Digital replica consent under SAG-AFTRA has to come with a "reasonably specific description of the intended use." Meaning "we'll use your voice for promos" is too vague, but "we'll use your digital replica for 2x 30-second trailers on streaming and social, in English only, for 12 months" is the kind of detail they're looking for.
What California adds on top
California AB 2602 actor consent rules say: any clause that swaps a human performance for a digital replica without a clear list of uses and proper representation can be treated as against public policy. To count as informed consent, the performer either has legal counsel negotiating the digital replica terms, or a union contract that already covers these AI uses.
AB 1836 applies the same logic to deceased performers. If you want to resurrect a voice for a new film or series, the estate has to actively agree to that specific AI use of the performer’s voice or likeness, with the same kind of clarity around scope and context.
Non‑union work and NAVA
In the non-union space, the National Association of Voice Actors (NAVA) has published AI Voice Agreement and Rider templates that reflect emerging best practices:
-
clear scope of use
-
limits on sensitive content
-
data handling rules
-
mechanisms for revocation
These documents are not binding law, but they are increasingly treated as a reference standard for ethical, defensible AI voice agreements.
Respeecher requires explicit written consent from the voice owner for every project. Without a signed, project‑specific agreement, we simply don’t start any digital replica work. Learn more →
What a Proper AI Voice Consent Agreement Must Include
A serious AI voice cloning consent agreement spells out what the system will do, where the voice can appear, and when that permission ends. Real actor consent for AI voice use depends on documented Consent Scope, a simple Right of Revocation, and a separate Training Data Clause that covers model use instead of hiding it.
|
Agreement term |
What the agreement should say |
|
Technology description |
State exactly what is being done with the voice: voice cloning (speech-to-speech, TTS) or another workflow. Avoid umbrella wording that hides the process. |
|
Intended use |
Name the project, platform, territory, language, and content type so the actor sees where their voice will actually show up. |
|
Consent Scope |
Define where, how, and for how long the voice can be used — the practical boundary line of the permission being granted. |
|
Volume and term |
Specify how much audio is being generated, how long the permission lasts, and whether renewal or extension is possible. |
|
Compensation |
Spell out pay for AI voice work clearly — session fees, reuse, and any extra rates for additional projects, union or non‑union. |
|
Prohibited uses |
List excluded contexts such as political ads, adult content, controversial endorsements, or unrelated future productions. |
|
Right of Revocation |
Explain whether the actor can withdraw consent, under what conditions, and what happens to already generated material. |
|
Training Data Clause |
State separately whether the voice may be used to train models, improve systems, or create future synthetic assets. This should never be implied. |
|
Explicit Written Consent |
Make consent clear, separate, and signed or initialed where required — especially for union‑covered digital replica work. |
|
Estate Consent |
If the performer has passed away, identify who is authorized to approve the use and for which project. |
How to Have the Conversation Without Losing the Actor
If you want genuine actor consent for AI voice, make it part of casting:
-
Bring up your AI voice cloning consent agreement when you talk about the role and schedule, so AI feels like a normal production tool.
-
Explain the use case in concrete terms: “we’ll use AI to record pickup lines in post so you don’t have to come back to the studio” lands much better than “we’re going to clone your voice.”
-
Loop in the agent or lawyer from the beginning; it shows you’re taking voice cloning permission seriously and reduces the sense that anything is being hidden.
-
Offer a review step: give the actor or their representative the right to listen to AI‑generated lines before final approval. This alone resolves a large chunk of objections.
-
Avoid broad contract language like “throughout the universe and in perpetuity”; in states with AB 2602, that style of wording is exactly what gets AI consent challenged.
How Respeecher Handles Consent — and Why It Matters for Studios
Respeecher starts every project with Explicit Written Consent from the voice owner, so actor consent for AI voice is the entry gate. If there’s no signed AI voice cloning consent agreement, the project doesn’t move — union, non‑union, or estate, the rule is the same for everyone.
For studios, that makes the provider a built‑in safety net. Respeecher won’t touch a brief without clear voice cloning permission, which keeps work inside Digital Replica (SAG-AFTRA) rules and new laws like AB 2602, instead of skating along the edge.
Add a public ethics policy and partnerships with Partnership on AI and Adobe's Content Authenticity Initiative, and consent stops being "a PDF somewhere" and becomes part of the way synthetic audio is produced and tracked.
How posthumous projects stay ethical
Posthumous work plays by the same rules, with more paperwork and more feelings on top. Respeecher only moves ahead on the basis of Estate Consent, with signatures from families or foundations before anyone touches a voice model, as in the Tommy Muñiz, Riky Rick, and Wilt Chamberlain projects.
The hidden bonus for studios: when you can point to that consent trail, actor AI voice rights stop being a scary abstract topic and turn into something very normal and very useful — documentation your lawyers can actually sleep on.
Final Thoughts
If AI keeps blowing up deals, it's usually because the consent conversation arrives like a surprise twist instead of page one of the script. Start it early, define the edges, and "we're using AI" reads like a production note rather than a plot twist.
Once you do that, the right provider makes the whole thing feel pleasantly dull. If consent is wired into every job, actor AI voice rights become part of the production log. Respeecher leans into exactly that: Emmy-level audio, public ethics, and a clear consent record on each project, so you spend less time firefighting and more time making the show. Learn more →
.png?width=477&height=264&name=How%20Much%20Does%20ADR%20Cost%20(1).png)
