A digital presenter is not simply a video generator with a portrait prompt. It is a reusable likeness and performance asset with a different risk profile from a fictional character. HeyGen’s September 2026 Avatar V release introduced a workflow that derives motion and expression from a short video, then applies a personal performance style to a photo-based appearance. The central production questions are who granted permission, how the avatar is governed and what gets billed each time it speaks.
What the product describes
HeyGen announced Avatar V on 9 September 2026, describing capture from a fifteen-second reference video. Its help page lists Avatar V’s Video Look usage at 48 credits per generated minute. Its broader credit documentation distinguishes Avatar III, IV and V and warns that credit consumption can depend on engine and generation type. A one-minute presenter output and a one-minute live-shoot source are different quantities: budget the video output and avatar-training costs independently.
HeyGen’s digital-twin consent instructions, updated in April 2026, require a separate consent video for each video-based digital twin. When creating an avatar of another person, that person must record their own consent. The provider checks the identity against source footage. This is a platform requirement in addition to, not a replacement for, releases and applicable privacy or publicity rights.
Record and approve one persona
Use evenly lit footage with a visible face, intelligible speech and consistent appearance. Before recording, agree the intended languages, scripts, permitted distribution channels, retention period and authority to revoke or suspend future generation. Store a non-public release register and avoid sharing raw consent files beyond authorized operators. Review an initial short sample for incorrect facial motions, false emotional emphasis and pronunciation before bulk production.
For an enterprise spokesperson, appoint a brand owner who can approve new scripts and a technical owner who manages account access. Human approval should be required for statements about medical matters, finance, legal obligations and changes in company policy. For marketing, verify claims in the actual script and ensure the avatar does not imply an endorsement the subject has never given.
Production controls and limitations
A synthetic likeness may produce strong eye contact but can still mis-handle gestures, visible objects, teeth and abrupt head motion. Avoid switching a live portrait between inconsistent looks without checking viewer confusion. Keep a clear distinction between visual performance generation, voice cloning permission, translated scripts and final editorial subtitling. The existence of a consent workflow is not proof that any uploaded celebrity or employee photo is authorized.
Cross-team boundary: dedicated voice model features belong to AI 05; this article addresses the presenter-video engine, consent and approval workflow. Recheck live regional access, enterprise controls and plan entitlements before promising an avatar slot or commercial-use term.
Worked production example
An instructor wants reusable video lessons recorded under their authorized digital identity. Before training a likeness, write the boundaries: internal employee education, named languages and no celebrity endorsements. Capture the provider’s required consent recording and keep an approval contact for new scripts. Generate a small sample containing a technical term, a smile and a hand movement; look for unnatural gestures or false emphasis. If a script later changes the instructor’s professional position, require fresh human approval even though the avatar could produce it without that approval.
Relevant alternatives and trade-offs
For simple non-visible narration use audio-only localization and conventional subtitles; for a separately consented enterprise presenter examine Synthesia or a live shoot.