Synthesia serves workflows in which a consistent presenter and repeatable instructional structure matter more than cinematic spectacle. A personal avatar can represent an authorized speaker across training material, while dubbing modifies an existing video into another language. These are separate production pathways with separate evidence and consent obligations.
Creating a photo-based personal avatar
Synthesia’s 20 July 2026 guide says Personal Avatars from photo are available on paid plans, with listed allowances of three for Starter, five for Creator and unlimited for Enterprise. Its most important rule is not the numerical allowance: the service requires a live consent video from the same person as the photo. It prohibits avatars of public figures or anyone without explicit consent. The person must also meet applicable legal age requirements.
For a training program, photograph the authorized presenter in a controlled environment, create the required consent asset and agree in writing which team may generate future scripts. Approve the accent, facial look, clothing and pacing for one short pilot scene. Keep that first review sample as a style reference, not as an assumption that every future script will render identically.
Dubbing an existing presentation
Synthesia’s dubbing guide describes uploading an existing video or providing a supported video link, choosing target languages and optionally enabling lip sync. It also offers adaptive duration settings to help align speech with translated timing. Longer words in a target language can change segment duration; such an option reduces timing conflict but does not guarantee that slide transitions, animations or legal disclaimers remain in sync.
Before dubbing, correct the original transcript and ensure the instructional content is current. Localize units, customer-support information, laws and dates as required. Once the dub is returned, have a qualified reviewer inspect terms, numbers and screen cues. If the video includes visible faces, inspect close-ups for mouth and jaw artifacts. Create localized subtitles separately if audience accessibility requires them.
Organizational guardrails
Separate ‘avatar exists’ from ‘any employee may use it’. Record a named approver, permitted channels and expiration or revocation rules. If an employee leaves or requests removal, have a documented suspension process for future synthesis. Be explicit to viewers where a digital instructor is used, especially when the material could imply live endorsement or a real interview.
The listed allowances are dated vendor terms, not universal regional guarantees. Recheck subscription entitlements in the target account; do not invent maximum output resolution, queue times, commercial-rights clauses or exact per-minute costs where not established in the primary pages. Voice and linguistic research beyond the video workflow is handed to AI 05.
Worked production example
A company has a consented trainer who appears in many course modules. Build one personal avatar from an approved photo and consent video, then render a pilot with a slide transition and a difficult pronunciation. For dubbing an older course, decide whether adaptive timing would shift screen events beyond their correct moment. If the German narration reaches a diagram later than the English original, retime the slide or adjust text; do not simply approve smooth-looking lips. Retain a human owner who can withdraw or revise future avatar content.
Relevant alternatives and trade-offs
For localized camera-original footage inspect HeyGen dubbing or a traditional language editor; for cinematic B-roll use a shot-generation tool rather than a talking-avatar product.