Case Study: MARZ accelerates VFX workflows with Weights & Biases

Leveraging AI for Visual FX at MARZ

MARZ, a visual effects (VFX) studio, faced the challenge of maintaining visual continuity in film and television, a process that consumed a significant portion of their budget on tedious, manual tasks like fixing visible wig lines. Their goal was to build AI tools to speed up these cumbersome workflows, allowing their artists to focus on creative work instead. To achieve this, they sought to leverage their vast repository of past work as training data for machine learning models.

The solution involved using Weights & Biases, specifically W&B Tables and W&B Reports, to evaluate, analyze, and track the performance of their ML experiments. This provided the team with vital insights, enabling them to quickly identify issues and share findings across the entire team. By implementing Weights & Biases, MARZ successfully developed their first ML product, Vanity AI, which performs VFX tasks 300 times faster than traditional pipelines, making the process significantly more cost-effective and freeing artists from capacity constraints.

MARZ

Thomas Davies

Research Team Lead