Avatars, digital twins and synthetic presenters

Some AI tools go beyond helping a creator make content and start standing in for the creator. An avatar tool can turn a script into a video of a realistic presenter speaking it. A digital twin is a model trained on one real person's face, voice and mannerisms, able to produce new video of that person saying things they never recorded. Virtual influencers are entirely synthetic characters with their own names, personalities and followings. These technologies can multiply a creator's output, enable new formats and make content in many languages, but they also raise the deepest questions in this course about consent, authenticity and trust.

Three kinds of synthetic presenter

• Stock avatars: realistic presenters supplied by a tool provider, usually based on paid actors who agreed to have their likeness licensed. They are commonly used for training videos, explainers and corporate content.
• Digital twins: avatars built from a specific real person, often the creator, trained on recorded video and voice samples. They can deliver new scripts in the person's likeness and, combined with dubbing, in other languages.
• Virtual characters: fully synthetic personalities, from cartoon mascots to photorealistic virtual influencers, that may be voiced and written by a team and sometimes animated with performance capture.

Each raises different issues. Stock avatars depend on the original actors' licences. Digital twins depend on the consent and control of the person they copy. Virtual characters raise questions of disclosure: audiences may not realize a realistic influencer is not a real person, or that a team or a brand is behind them.

What creators use them for

Creators and media companies use synthetic presenters in several ways:

• Scaling evergreen content: turning written guides into short video explainers without filming each one.
• Localization: having a digital twin deliver content in languages the creator does not speak.
• Corrections and pick-ups: fixing a line, updating a price or replacing an outdated fact without reshooting.
• Personalized messages: generating many short greetings or thank-you videos for supporters.
• Characters and mascots: running a branded character across platforms, sometimes with live interaction.
• Accessibility and continuity: allowing a creator to keep publishing during illness or recovery, with clear disclosure.

Not every use is wise. Audiences often follow creators because of real, unscripted presence. Replacing that with a twin, especially without saying so, can feel like a betrayal when discovered. Many creators limit twins to clearly labelled, practical uses, such as translations or corrections, and keep personal and opinion content human.

Consent and control for digital twins

A digital twin of yourself is a powerful asset and a serious risk. If it is misused, leaked or controlled by someone else, it can make you appear to say or endorse anything. Before creating one, consider:

• Who owns and controls the model: you, the tool provider, your company or a business partner?
• What the provider may do with your face and voice data, including training other models, and whether you can delete them.
• Who on your team may generate content with the twin, and what approval is required before publishing.
• What happens to the twin if you leave a company, sell your business, or die.
• How you will mark twin-generated content so audiences and platforms can tell it apart.

For any digital twin of another person, such as a co-host, presenter or employee, written consent is essential and should cover the same points the lesson 5 voice consent covered: purpose, scope, duration, payment, approval rights, withdrawal and deletion. Employment and contractor agreements should state clearly whether the company may create or use a person's likeness after the working relationship ends; many jurisdictions now treat broad, vague grants of digital replica rights with suspicion, as lesson 10 explains.

Virtual influencers and disclosure

Virtual influencers can build large followings and attract brand deals. The disclosure questions are layered. First, is the audience aware the character is not a real person? Second, if the character promotes a product, is the material connection with the brand disclosed? Third, who is actually speaking: a creative team, an agency, or the brand itself?

Advertising regulators in both countries apply the same endorsement rules to virtual and AI influencers as to human ones. In Canada, the Ad Standards Influencer Marketing Disclosure Guidelines (updated October 2025) state that AI-generated influencers must follow the same material connection disclosure rules, and recommend as best practice disclosing that the spokesperson is virtual, using tags such as #VirtualInfluencer or #AIinfluencer. In the United States, the FTC Endorsement Guides apply to endorsements regardless of the technology used, and the FTC has warned that a virtual influencer cannot honestly claim personal experience with a product it cannot use.

The economics of a digital twin

Digital twins promise to cut production time, but they come with their own costs: subscription or per-minute generation fees, script writing, review and editing, and the risk of lower audience engagement. A simple comparison helps decide where they make sense.

Live and interactive avatars

Some tools let an avatar or AI character respond to viewers in real time, in live streams, chats or comment replies. Interactive avatars raise further issues: the AI may say things the creator would never say, give harmful advice, form parasocial bonds with lonely or young viewers, or be manipulated by users into offensive statements. If you deploy an interactive character, set clear limits on topics, monitor it, make its AI nature obvious, provide a way to reach a human and keep logs. Lesson 7 covers AI in community management more generally.

Deceased performers and legacy likenesses

Technology can now recreate people who have died, from historical figures to creators' own relatives or late collaborators. Laws in some places protect the likeness of deceased people for a period after death, and families and audiences can react strongly to unexpected recreations. Avoid recreating a real deceased person without the permission of whoever controls their estate or likeness rights, and be cautious and transparent even when you have permission.

A digital twin agreement checklist

Whether you are creating a twin of yourself with a vendor or of a team member for your company, put the terms in writing. A practical checklist for the agreement includes:

• Parties and purpose: who is creating the twin, for what kinds of content, and on which platforms.
• Scope: languages, topics that are allowed and topics that are excluded, such as political statements, sponsored endorsements or personal messages.
• Approval: who must approve each script and finished video before publishing.
• Payment: a fee per use, per minute generated or a flat licence, and how it changes over time.
• Duration and termination: how long the licence lasts and what happens to existing content when it ends.
• Data and security: where face and voice data are stored, who can access them, and whether they can be used for training other models.
• Withdrawal and deletion: how the person can withdraw consent and confirm the model and data are deleted.
• Disclosure: how twin content will be labelled to audiences.
• Succession: what happens if the person dies, becomes unable to work or the business is sold.

Some creators keep their twin model with a vendor under an account they control, rather than one controlled by an agency or network, so they can move or delete it later.

Audience reactions and testing

Before relying on a twin, test audience reactions. Publish a small number of clearly labelled twin videos, read the comments, compare retention with similar human-presented videos and ask your community directly. Some audiences find twins useful for translations and quick updates; others feel distanced. The data, not the tool's marketing, should decide how far to go.

Labelling twin content clearly

Good labelling for twin content uses several layers: the platform's AI label, a short line in the description such as "This video is presented by my AI digital twin, from a script I wrote and approved", and, for longer videos, a brief on-screen note at the start. Being specific about what is human, such as the script and the approval, helps audiences understand that the creator is still responsible for the content.