Governance, Ethics & Legal Practical insights
Synthetic presenters need a clear permission brief
Plan synthetic marketing presenters with documented likeness and voice permissions, accurate scripts and clear disclosure decisions before production.
A synthetic presenter can make localized videos easier to produce, but it also creates a new kind of production brief. The team must define whose appearance and voice may be used, what the presenter may say and what the audience is likely to believe. Solve those questions before generating variants.
Specify the permitted performance
For a real person’s digital likeness, ask the rights owner and legal reviewer to define the permitted channels, territories, languages, campaign subjects, term and approval process. Address future reuse and ending the arrangement. A release for filming a webinar should not be treated as proof that unlimited synthetic performances were authorized.
Keep the script within the evidence
Review every language version against the approved product facts. A presenter must not imply personal product experience that nobody actually had. If an avatar explains a service, introduce it as a presenter rather than dressing it as an independent satisfied customer. Separate any genuine customer quotation from the generated performance.
Decide what viewers need to understand
Assess whether the output falls within the relevant deepfake definition and disclosure obligations. Article 50 distinguishes deployer disclosure from provider marking; creative context can affect the applicable presentation of disclosure. [2] The voluntary transparency code offers an implementation framework, not permission to use a person’s identity. [1]
Approve the rendered result
Watch the actual video with sound on and off. Check lip synchronization, pronunciation, visible notices and whether the generated delivery changes the meaning of a qualification. Keep the signed permission scope, final script, rendered file and approval together. If the voice or face changes later, reopen the review instead of assuming the original approval covers a different performance.
Sources and evidence
Sources checked on 4 October 2026. Proposed workflows and hypothetical examples are editorial analysis.
From insight to practice