Phantom Subject2Video Run Locally In ComfyUI - Better Then Wan 2.1 VACE?
Added 2025-04-28 13:34:51 +0000 UTC
Phantom ComfyUI Tutorial : https://youtu.be/2toaC3-yDcY
Unlock the power of ByteDance’s Phantom AI—the revolutionary subject-consistent video generation framework! In this video, we dive deep into Phantom’s ability to create seamless, multi-subject videos using WAN 2.1 as its base model. Watch as we test its performance in ComfyUI, compare it to WAN V-Ace, and reveal how it preserves character consistency across scenes—even with mixed styles like anime and realism. Plus, get pro tips on optimizing VRAM usage, workflow setups, and how to leverage multiple reference images for storytelling.
Who Is This For?
AI video creators wanting character-consistent storytelling.
ComfyUI users exploring Phantom’s WAN 2.1 integration.
Filmmakers and animators needing multi-subject video synthesis.
Tech enthusiasts tracking ByteDance’s latest AI breakthroughs.
Why It Matters:
Phantom bridges the gap between static references and dynamic videos, offering unprecedented consistency for narratives, ads, and creative projects. While still evolving, its ability to track outfits, hairstyles, and poses across scenes makes it a game-changer for AI-driven content.
Resources:
bytedance-research/Phantom
https://huggingface.co/bytedance-research/Phantom
For ComfyUI:
git clone https://github.com/kijai/ComfyUI-WanVideoWrapper.git
for update :
In "ComfyUI-WanVideoWrapper" folder
use CMD : git pull
Diffusion Model:
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Phantom-Wan-1_3B_fp32.safetensors
or
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Phantom-Wan-1_3B_fp16.safetensors
Attached workflow that I created in this tutorial: