The ComfyUI HunyuanCustom Multi-Subject workflow integrates Tencent's advanced HunyuanCustom model into the ComfyUI environment, specifically optimized for dual-subject generation. Built upon the HunyuanVideo framework, this HunyuanCustom workflow focuses on generating videos featuring interactions between two subjects (such as a person and a product) with exceptional identity preservation. The HunyuanCustom workflow, optimized by Kijai, includes several enhancements like FreSca technology and specialized nodes designed to improve HunyuanCustom generation quality, consistency, and efficiency.
Example Setup for HunyuanCustom Multi-Subject:
Run
to run the HunyuanCustom workflowMain HunyuanCustom Node:
Resolution
: Sets HunyuanCustom output video resolution (recommended: 720×1280 for best quality)Frames
: Number of frames to generate in HunyuanCustom videos (typically 129 frames for 5-second videos)Steps
: Number of denoising steps for HunyuanCustom (30 recommended for balance of quality and speed)CFG Scale
: Controls prompt adherence strength in HunyuanCustom (higher = more prompt-aligned but potentially less natural)Flow Shift
: Adjusts motion fluidity in HunyuanCustom outputs (recommended value: 13.0)FreSca Enhancement Node:
Enabled
: Toggle the FreSca enhancement on/off for HunyuanCustomScale
: Strength of FreSca effect (1.0 recommended for balanced enhancement)Mode
: Set to "single blocks" for optimal hand stability and detail preservationID Enhancement Module:
Weight
: Controls how strongly the identity features are preserved in HunyuanCustomTemporal Concat
: Enables improved identity consistency across framesSLG (Stability-Latent-Guidance) Node:
Mode
: Set to "single blocks" for best hand detail preservation in HunyuanCustomStrength
: Controls how strongly the stability guidance is applied to the generationBest Practices for Subject Images:
Improving Hand Rendering in HunyuanCustom:
Performance Enhancement for HunyuanCustom:
Quality Enhancements:
For additional details and development references:
This workflow is powered by HunyuanCustom, developed by Tencent's Hunyuan team. The original paper "HunyuanCustom: A Multimodal-Driven Architecture for Customized Video Generation" introduced this powerful framework that emphasizes subject consistency while supporting image, audio, video, and text conditions.
The ComfyUI integration of HunyuanCustom is provided by Kijai, including optimizations like FreSca technology, SLG nodes, and TCatch implementation that significantly improve generation quality and efficiency. Full credit goes to the original authors at Tencent for their groundbreaking work on HunyuanCustom and to Kijai for the outstanding ComfyUI implementation and enhancements.
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Playground, enabling artists to harness the latest AI tools to create incredible art.