ByteDance Open-Sources Bernini: Unified AI Video Editor
ByteDance has released Bernini as open source — a unified video editing model that accepts text, image, and video reference inputs to perform: character addition and removal, tone and perspective changes, background and season replacement, orbit-style video generation from multi-angle stills, and multi-image character consistency. Based on a 1.2 architecture with high- and low-fidelity branches totalling approximately 168GB. ComfyUI integration is actively underway.
Why It Matters
Bernini is the first open-source model to match the semantic editing scope of closed video-editing tools — making production-grade character-consistent video editing accessible without API costs, directly relevant to automated content production pipelines.