Deepfake technology has become a practical visual effects tool in film and television, used to de-age performers, build digital doubles, and even recreate deceased actors for new scenes. The same capability that enables those effects also creates real ethical exposure: AI can produce "hyper-realistic videos of people doing or saying things they never actually did," as VFX practitioner Marc Czerwiec of Light-Based Studio notes, and that dual nature is exactly what productions must manage. This guide explains what deepfakes in VFX actually do, then walks through a decision framework covering consent, authenticity, and misuse that a producer or VFX supervisor can apply before committing to the technology. It is a guide for consideration, not legal advice, and not a technical tutorial.
Understanding Deepfakes in Visual Effects
In a production context, a deepfake is an AI-generated or AI-assisted synthetic likeness—typically a face or voice swapped onto footage of another performer or a digital model. Czerwiec, writing on generative AI in visual effects, describes the core capability: machine-learning systems that generate highly convincing digital humans, which studios can deploy for shots that would otherwise be expensive, impossible, or physically risky to capture practically.
What makes the technique distinct from traditional CGI is that it learns from recorded footage of a real person. That is both its creative strength and its ethical weight: the output is tied to a specific, identifiable performer whose likeness is the raw material. Analysis of the film industry's adoption of deepfakes between 2017 and 2024 by Bulgarova, Ullah, Mondal, and Ma concludes that as the technology evolves, "the need to develop ethical principles and a regulatory framework will become more and more apparent"—a conclusion echoed by nearly every industry and academic commentary on the subject.
For filmmakers, the practical takeaway is that the technology itself is neutral and increasingly accessible, but every use inherits obligations toward the person being replicated and toward the audience watching the result. Those obligations are the substance of the framework later in this article.
Key Creative Applications of Deepfake Technology
Czerwiec identifies the two signature production uses: creating digital doubles of actors and de-aging performers. De-aging uses the technique to make an actor appear younger for flashback scenes or long-running franchises, letting a single performer carry a character across decades without recasting or heavy conventional makeup effects.
Digital doubles extend the same machinery to stunt work and extended performances. Instead of a face replacement composited by hand, the AI is trained on the actor's likeness and maps it onto a stunt performer, producing shots that are realistic enough to pass as the actor while keeping the human performer out of danger.
The most contentious application is the posthumous performance: using deepfake methods to recreate a deceased performer for new material. Czerwiec notes that "deepfake technology can be used responsibly in film production—such as creating digital doubles for deceased actors"—but the practice raises unresolved questions about legacy rights and who is entitled to authorize a dead performer's likeness.
How close the results sit to indistinguishable-from-real is illustrated by the widely covered Tom Cruise parodies. NBCU Academy reported that the deepfake Cruise, built around performer Miles Fisher by the company Metaphysic, was convincing enough that Fisher and Metaphysic began selling deepfake creations commercially. Fisher framed a consented version of the practice in business terms: "Let's say Tom Cruise gave us the consent for his likeness where we could move beyond just small parody clips… Everybody gets paid for that intellectual property." Metaphysic says it takes on only projects it considers ethical, though NBCU Academy also noted that some government officials worry that leaving such decisions to companies and individuals sets a dangerous precedent.
Ethical Considerations: Navigating the Deepfake Landscape
Three themes recur across industry and academic commentary: consent, authenticity, and misuse. Consent is the clearest flashpoint. Writing on the ethics of generative AI for Carnegie Mellon's Tepperspectives, Nick Wright argues that as AI enables easier and more convincing fabrications, it "highlights the need for stronger ethical and regulatory frameworks to protect individuals from misuse and misrepresentation." His framing—drawn from cases like voice cloning—treats a person's likeness and voice as identity that cannot simply be borrowed for convenience.
For deceased performers, consent becomes a question of legacy and estate rights. SAG-AFTRA has championed legislation requiring consent for the use of dead performers' likenesses in AI-created digital replicas, an effort aimed at closing the gap where a performer can no longer speak for themselves.
Authenticity is the second concern. If audiences cannot tell synthetic footage from real performance, productions gain power over what viewers believe about real people. Lucianne Zammit, writing for the 3CL Foundation, argues that laws and policies must balance protecting individuals' rights and privacy against freedom of expression and innovation, and that "establishing clear legal standards for the use of deepfake technology, including consent requirements and penalties for malicious use, can help deter misuse while supporting beneficial applications."
Third is misuse beyond the production itself: material created for a film can circulate stripped of context. Zammit adds that detection technologies form "a continuous arms race against evolving deepfake generation techniques," meaning productions cannot rely on outside systems to contain what they release. A research review published in the International Journal of Research and Review summarizes the consensus: collaboration between technologists, artists, ethicists, and policymakers will be crucial to harnessing deepfakes' benefits in VFX while mitigating risks.
Ethical Decision Framework for Deepfakes in VFX
Use the matrix below during pre-production to evaluate a proposed deepfake use against the criteria the evidence identifies as decisive: the status of consent, the transparency of the effect, and the risk of malicious re-use. Work through it before shots are budgeted, because consent and disclosure obligations change the cost and feasibility of a sequence, not just its ethics.
| Dimension | What to establish | Key question for your project |
|---|---|---|
| Creative goal | Classify the application: de-aging, digital double for stunts, posthumous performance, or voice cloning. | Is the deepfake the least-invasive way to achieve this shot, or is a practical alternative available? |
| Consent — living subject | Explicit, documented permission from the performer for this specific use; Wright's analysis treats unconsented likeness use as misrepresentation of identity. | Is there a signed agreement covering this exact use, scope, and compensation? |
| Consent — deceased subject | Authorization from the estate, informed by the legislative push SAG-AFTRA has championed for consent over dead performers' AI replicas. | Who has the legal and moral standing to say yes, and is their authorization in writing? |
| Authenticity and disclosure | A plan for audience transparency about synthetic elements, mitigating misrepresentation risks Zammit and Czerwiec both identify. | Will the audience know this performance was synthesized, and does the answer matter to your film's claims? |
| Potential for misuse | Assessment of whether the footage, taken out of context, could show the person doing or saying things they never did. | If this clip circulated without context, who could be harmed and how would you respond? |
Apply it as a sequence, not a checklist to skim. First fix the creative goal, because it determines which consent regime applies. Second, resolve consent before any asset is built—Fisher's Cruise scenario only works ethically in his own telling because the likeness is consented and paid for. Third, decide on disclosure. Fourth, run the misuse test, including whether out-of-context circulation could plausibly deceive viewers about the real person. A project that cannot pass a stage should not advance until it can.
The framework is a guide for consideration, not a definitive or exhaustive legal compliance checklist; productions with real exposure should seek qualified counsel.
Making Responsible Choices in Deepfake VFX
Filmmakers and VFX professionals should adopt a proactive ethical review process for any project involving deepfake technology, prioritizing explicit consent, transparency, and careful consideration of potential misuse. The evidence points consistently in that direction: consent is the dividing line between the responsible uses Czerwiec describes and the misrepresentation Wright warns against, while Zammit's analysis shows the legal environment moving toward mandatory consent requirements rather than self-policing.
The measurable indicator of a responsible process is the presence of a documented consent agreement and a clear disclosure plan for deepfake elements in production documentation. If a project's paperwork shows both, the production has made its ethical position verifiable rather than aspirational—if either is missing, that gap is the next concrete milestone to close before cameras roll.
Sources
- Exploring the Ethical Implications of genAI in Visual Effects: — Light-Based Studio — Light-Based Studio | VFX, Film & Data Solutions
- Bulgarova B.A., Ullah Md T., Mondal P., Ma F. From Emergence to Ethical Frontiers
- 'Deepfake' Technology and Ethics — NBCU Academy
- Deepfakes and the Ethics of Generative AI — Tepperspectives
- QUORA
- Beyond Reality: Navigating the Ethical Minefield of Deepfake Technologies - 3CL Foundation
- IJRPR
