AI Video Pipeline Architect - ComfyUI

ComfyUI Workflow Expert Needed: SAM 3 + Qwen + Wan 2.1 Video Pipeline

US AI Video Agency is looking for a ComfyUI architect who is already working with the bleeding edge of open-source AI. I need a robust, automated workflow replicating the Meta SAM 3 (SAM 2.1) -> Qwen Image Edit -> Wan 2.1 pipeline for high-fidelity virtual try-on and video object replacement.

The Mission:
Segmentation: Implement SAM 3/2.1 to generate precise, temporal masks for video (handling Object IDs and occlusion correctly).
In-Painting: Integrate Qwen Image Edit for reference-based clothing/object replacement on the base frame.
Video Synthesis: Pipe the masks and reference edits into Wan 2.1 (Video-to-Video) to animate the change consistently across all frames.
Optimization: Ensure the workflow handles VRAM efficiently and includes an upscaling pass (Z-Image Turbo/RealESRGAN).

You Must Have:
Deep expertise in ComfyUI (custom node management, latent masking, bounding box logic).
Working knowledge of SAM 3 / SAM 2.1 (specifically Object ID propagation).
Experience tuning Wan 2.1 and Qwen parameters for consistency.
Ability to troubleshoot Python/PyTorch dependency conflicts if they arise.

 ... Show more