Innovate Futures @ Benji
patreon
Wan 2.1 Base MiniMax-Remover And NAG for Video Object Removal - Handy VFX AI Tool!
🕑 Added 2025-06-21 14:41:37 +0000 UTC
Video : https://youtu.be/KqkdKtx91SM
Explore the power of Wan 2.1 with MiniMax-Remover and NAG (Normalized Attention Guidance) for advanced video object removal in this tutorial! Learn how this AI tool leverages diffusion transformer models to trim unwanted objects, shadows, and noise from videos while maintaining natural results. Discover practical workflows, testing insights, and refinements using ComfyUI, ControlNet, and Fusion X for higher-quality outputs. Whether you're a VFX artist, video editor, or AI enthusiast, this tool offers a cost-effective solution for object removal—despite its limitations with larger objects and shadows.
Who is this content suitable for?
VFX professionals, video editors, AI researchers, content creators, and anyone interested in AI-powered video post-production tools.
Why it matters:
The MiniMax-Remover and NAG combo provides an accessible way to clean up videos by removing unwanted objects, but it also highlights the current limitations of smaller AI models (1.3B parameters). This video showcases real-world testing, refinements, and workarounds to achieve better results, making it invaluable for those working with AI-driven video editing.
MiniMax-Remover Project : https://github.com/zibojia/MiniMax-Remover
Model: Wan2_1-MiniMaxRemover_1_3B_fp16.safetensors
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-MiniMaxRemover_1_3B_fp16.safetensors
NAG
https://chendaryen.github.io/NAG.github.io/
Attached tutorial workflows below
