Vmake Integrates ICML-Accepted BridgeRemoval Algorithm for Video Object Removal

Vmake has integrated the BridgeRemoval algorithm into its AI video object removal system, announced in a press release. The algorithm, developed from an ICML 2026 accepted research paper, is designed to improve both the accuracy and speed of removing objects, watermarks, and overlays from videos.

BridgeRemoval introduces a stochastic bridging path that connects the original source video to the target output. This approach maintains consistent motion across frames and reduces flickering, a common issue in traditional diffusion-based video removal methods. The algorithm also processes video faster while preserving realistic background details.

The paper behind BridgeRemoval was accepted at the 43rd International Conference on Machine Learning (ICML) in Seoul, confirming the method’s technical merit and originality through peer review. The integration allows Vmake’s video watermark removal feature to support multi-object and batch removal, improving workflow efficiency for creators and small teams.

Vmake’s updated tool enables users to remove watermarks, logos, text, or people from video scenes with precise targeting. The system restores backgrounds smoothly, keeping the final visuals natural and stable.