Identity first
Protect facial features and keep the effect below the neck wherever possible.
A repeatable AI video transformation pipeline for a major sports event in Belgrade, built in Veo 3.1 to scale across player identities while protecting recognizable faces and a consistent visual progression.

For the European Water Polo Championship in Belgrade, I developed and executed a custom AI video transformation workflow in Veo 3.1. The concept used progressive player transformations that evolved as teams advanced through the tournament.
Rather than treating every clip as a one-off generation, I shaped a repeatable process that could be applied to a large player pool while maintaining identity, believable motion and a shared visual language.
The transformation needed to focus on the body while preserving each player's head and recognizable identity. Armor concepts from the art team had to be adapted to different source images and body types, with clear home and away variants and a progression that read immediately from one tournament stage to the next.
Protect facial features and keep the effect below the neck wherever possible.
Build from a first line structure through multiple armor states to an enhanced finalist form.
Translate armor drawings into usable targets that fit different body shapes and player references.
Keep the method practical across 16 teams, 15 players per team, team-color variants and stage changes.
A controlled line design builds over the torso and arms while the head, face and identity remain unchanged.
The line work becomes the foundation for armor elements that construct onto the player rather than simply replacing the source image.
The next tournament stage increases complexity and strength while staying continuous with the earlier design.
The finalist form reads as a meaningful upgrade without breaking the established visual language.
Source material and target placement were adjusted manually where needed. Quality checks focused on head preservation, facial identity, body-only effects, stage-to-stage progression, temporal continuity and whether each clip provided enough clean material for editorial use.
The deliverables were silent by client brief and upscaled to 1440 × 2560. The process combined reference-driven generation with art-direction adaptation, motion review and post-production thinking.
AI transformation breakdownVertical production reel · 1080 × 1920 source · silent by brief
This project demonstrates how I translate a visual brief into a scalable production system, work from art-direction input, preserve identity through complex transformations, and deliver a large, consistent asset set for a real event workflow.