AI digital human videos: where they help exporters, and how to make them responsibly.

What AI digital humans and AI video generation are

An AI digital human (数字人 in Chinese) is a realistic, computer-generated presenter whose face, voice and lip movements are produced by AI from a written script. Some are stock avatars licensed from a platform; others are custom avatars created, with permission, from recordings of a real person. AI video generation is the broader term for tools that create or edit video from text, images or existing footage.

For an exporter, the practical change is that the script, not the filming, becomes the main work. Change the text and you get a new version, in another language, without booking a studio or a presenter.

Where digital human videos help, and where they do not

Digital humans work best for factual content that must exist in several languages and change when the product changes. They work poorly where trust depends on seeing a real person or a real customer.

The testimonial point is a legal one in Australia, not only a matter of taste. The ACCC (Australian Competition and Consumer Commission) says it is unlawful to make a false or misleading representation that purports to be a testimonial by any person. An AI presenter posing as a satisfied customer risks being exactly that.

A production workflow that keeps videos accurate

A digital human video is only as accurate as its script, so the workflow should put product knowledge first and the avatar last.

1. Script from real product dataWork from approved specifications, data sheets and real customer questions. AI can draft, but every number must come from your records.
2. Product expert reviewAn engineer or product manager checks the script, and later the finished video, for accuracy and safety.
3. Choose the presenterUse a licensed stock avatar, or a custom avatar only with the person’s written consent.
4. Produce language versionsTranslate the approved script, then have a fluent speaker who knows the product check terms, units and pronunciation.
5. Disclose and labelSay in the video and its description that the presenter is AI-generated, and follow each platform’s labelling rules.
6. Publish, track and updatePlace the video on the relevant product page, track what viewers do next and remake it when the product changes.

Keep the approved script, each language version and the approval date together, so that when a specification changes you know exactly which videos to remake.

Disclosure: tell viewers the presenter is AI

Disclose clearly that a presenter is AI-generated. It is honest, it protects your reputation if a buyer finds out later, and in some markets and on some platforms it is required.

• Australia: the National AI Centre’s voluntary guidance, Being clear about AI-generated content, recommends labelling, watermarking and metadata to show when AI has been used to create or change content.
• China: the Measures for Labelling AI-Generated and Synthetic Content (人工智能生成合成内容标识办法), in force since 1 September 2025, require service providers to add visible labels to AI-generated video, and require users who publish such content to declare it and use the platform’s labelling tools.
• YouTube: creators must disclose realistic AI-generated or altered content when uploading, such as content that makes a real person appear to say something they did not say.

Rules differ by market and change often, so check the current rules in each country and on each platform where you publish. A clear on-screen line such as “AI-generated presenter”, plus a note in the description, is a sensible minimum.

Consent and likeness rights for a real person

If a digital human is based on a real person, such as your sales manager or a hired presenter, get their written consent before recording anything. A face and voice are part of a person’s identity, and a digital copy can be made to say things the person never said. The consent should cover:

• ScopeWhich videos, products, languages and channels the avatar may be used for.
• Duration and exitHow long the permission lasts, and what happens when an employee leaves or a contract ends.
• ApprovalWhether the person reviews scripts, and what they may refuse to say.
• Storage and deletionWhere recordings and the avatar are kept, who can access them and how they are deleted.
• Platform termsWhat the avatar platform’s terms say about ownership and whether your data may be used to train models.

In China, the Provisions on the Administration of Deep Synthesis of Internet Information Services (互联网信息服务深度合成管理规定) require services that offer face or voice editing to prompt users to inform the person being edited and obtain their separate consent. Rules elsewhere differ, so take legal advice for your markets. Licensed stock avatars avoid many of these questions, but still read the licence for commercial use and any restrictions.

How to measure digital human videos

Measure each video against the job it was made for. Views tell you little; what viewers do next tells you whether the video works.

• Product explainers: enquiries from the page, and how often buyers still ask questions the video already answers.
• Training and onboarding: time for new distributor staff to complete onboarding, and their assessment results.
• FAQ videos: the number of repeat support questions on the topics covered.
• Every video: completion rate by language, to spot a weak translation or a script that runs too long.

Videos work best on a site built to hold them; our export website guide covers product pages and enquiry tracking. Optivise AI’s Marketing & Lead Generation service produces company promo videos and custom digital humans from licensed likeness and voice, and our Global Foundation service builds the multilingual websites they sit on. Scope is agreed in each proposal.