02 / Production · AI
Five videos, two languages, no shoot
A repeatable avatar-video production system for product features and internal training.
- Client
- International corporate client
- Role
- Producer
- Period
- 2025
The brief
Useful videos, without a shoot per version.
Video explaining features of new products, plus educational material for staff. Traditional production means a shoot per piece, per language, per angle — impractical for the volume needed on the available timeline and budget.
Approach
A production system, not one-off assets.
Built as a production system rather than a set of one-off assets: scripting, avatar selection, voice, and iteration against feedback, structured so additional pieces cost incrementally instead of starting from zero each time.
Once the first avatar was agreed, the work moved into controlled refinement: correcting clothing, camera angles, hand gestures and lip-sync for both Romanian and Russian voice tracks. That meant treating the avatar as a production asset that could be directed and improved, not as a one-click output.
The same source material was covered from several angles rather than produced once and reused flat, so each piece served a distinct purpose rather than being a variant of the last.
Control system
Consistency across character, camera and gesture.
Two selected production sheets show the point of the exercise: Alex remains recognisable while posture, framing and delivery change from scene to scene.
Video sample
From a controlled character to a usable message.
A short production sample. It begins muted and loops automatically; use the control to pause or restart it.
Outcome
Ready for the next message.
A corporate-avatar production pipeline ready to create more internal-communication videos in Romanian and Russian, without returning to a physical shoot for every new subject, language or update.
Stack
HeyGen · ComfyUI · GPT · Final Cut Pro