How Outpost VFX Uses AWS to Accelerate AI Model Training for Visual Effects

This article describes how Outpost VFX accelerated AI model training for visual effects by implementing distributed multi-GPU training on AWS EC2 P5 instances, achieving 8x faster speeds.

• Outpost VFX faced single-GPU bottlenecks limiting face replacement model training to 1-2 weeks per iteration
• Migrated from local RTX 3090 GPUs to AWS EC2 P5 instances with NVIDIA H100 GPUs and NVLink interconnects
• Implemented PyTorch Distributed Data Parallel (DDP) training strategy with AWS Generative AI Innovation Center support
• Achieved 8x performance improvement, reducing initial client delivery from 1-2 weeks to 2 days
• Enabled training on higher-resolution images and larger datasets, improving output quality

The parallelized architecture provides Outpost VFX with scalable, secure AI-assisted face replacement capabilities integrated into their global VFX production pipeline.