<div data-youtube-video="https://youtu.be/UespHFT9cjc"><iframe src="https://www.youtube.com/embed/UespHFT9cjc" width="853" height="480" allowfullscreen></iframe></div> 
 
This workflow is the standard full-configuration version of the MiniMax H3 character replacement / reference editing route. It swaps the character in a reference video with the character defined by a reference image, keeps the original audio in the final clip, and is the baseline version of the three-variant swap set. 
 
The inspected graph uses the MiniMax H3 hybrid FL2VA/Ref2VA INT8 base model (minimax_h3_hybrid_fl2va_ref2va_zs05_b25-49_int8.safetensors), Qwen3-VL 32B NVFP4/AWQ text encoding, MiniMax H3 video VAE (INT8 ConvRot) and audio VAE (FP32), and OpenVDN DMD8 execution planning with custom samplers. Depth control comes from Video Depth Anything ViT-S (518x1280 fp16, gray), subject tracking comes from SeC full-clip segmentation with feathered mask compositing, and the swap is driven by the MiniMaxH3ReferenceToVideo control node that combines the Picture 1 identity reference with the single reference video depth composite (768x768, 362-frame setting). Reference audio is encoded through the LTXV audio VAE and mixed back into the final video; image presets cover 0.9MP and 0.48MP targets and the graph ships as a dual-instance configuration. 
 
This package is meant for ComfyUI and RunningHub users who want the complete swap route exactly as shipped in the episode: reference image identity, depth-composite reference video, audio preserved, and a single final export. It is the recommended starting point when evaluating the three variants of this episode. 
 
Main features: 
 
- Full-configuration character replacement with reference image identity plus reference video depth composite 
- Video Depth Anything ViT-S whole-frame depth and SeC full-clip segmentation with feathered compositing 
- MiniMax H3 hybrid FL2VA/Ref2VA INT8 base with OpenVDN DMD8 execution planning 
- Original reference audio encoded and mixed back into the final clip 
- RGB + depth person display preview output for compositing checks 
- Qwen3-VL 32B NVFP4/AWQ text encoding with MiniMax H3 video and audio VAEs 
- Dual-instance graph with 0.9MP / 0.48MP image presets and unified 24 FPS packaging 
 
Suggested workflow: 
 
Use this version as the baseline of the set: run the episode's reference image and video with the bundled prompt template, then judge whether the result already meets the goal. If facial detail or identity stability needs more work, move to the Face Refine variant with the same inputs; if audio is not required, use the Silent variant for faster iteration. 
 
For fair comparisons across the three variants of this episode, run one short test first with identical inputs, then review identity stability, motion consistency, and background drift before scaling up to full-length renders. 
 
RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
Workflow: https://www.runninghub.ai/zh-cn/post/2106048569508659202?inviteCode=rh-v1111 
 
If the results meet your expectations, you can later deploy it locally for customization. 
 
Fan Benefits: Register to get 1000 points + daily login 100 points - enjoy 4090 performance and 48 GB super power! 
 
Bilibili Updates (Mainland China &amp; Asia-Pacific) 
 
If you're in the Asia-Pacific region, you can watch the video below to see the workflow demonstration and creative breakdown. 
Bilibili Video: https://www.bilibili.com/video/BV1VjaD6jEU3/ 
 
Support Me on Ko-fi 
 
If you find my content helpful and want to support future creations, you can buy me a coffee. 
Every bit of support helps me keep creating. 
Ko-fi: https://ko-fi.com/aiksk 
 
Business Contact 
 
For collaboration or inquiries, please contact aiksk95 on WeChat. 
 
打开下方链接即可在线体验，无需安装。 
工作流：https://www.runninghub.ai/zh-cn/post/2106048569508659202?inviteCode=rh-v1111 
 
如果你觉得效果理想，也可以在本地进行自定义部署。 
 
粉丝福利：注册即可领取 1000 积分，每日登录再领 100 积分，体验 4090 和 48GB 大显存性能。 
 
B站视频（中国大陆及亚太地区） 
 
如果你在中国大陆或亚太地区，可以通过下面的视频查看工作流的实测效果与创作思路。 
B站视频：https://www.bilibili.com/video/BV1VjaD6jEU3/ 
 
我会在夸克网盘持续更新模型资源： 
https://pan.quark.cn/s/9ec095b95838

MiniMax H3 Role Swap Standard

<div data-youtube-video="https://youtu.be/UespHFT9cjc"></div> 
 
This workflow is the standard full-configuration version of the MiniMax H3 character replacement / reference editing route. It swaps the character in a reference video with the character defined by a reference image, keeps the original audio in the final clip, and is the baseline version of the three-variant swap set. 
 
The inspected graph uses the MiniMax H3 hybrid FL2VA/Ref2VA INT8 base model (minimax_h3_hybrid_fl2va_ref2va_zs05_b25-49_int8.safetensors), Qwen3-VL 32B NVFP4/AWQ text encoding, MiniMax H3 video VAE (INT8 ConvRot) and audio VAE (FP32), and OpenVDN DMD8 execution planning with custom samplers. Depth control comes from Video Depth Anything ViT-S (518x1280 fp16, gray), subject tracking comes from SeC full-clip segmentation with feathered mask compositing, and the swap is driven by the MiniMaxH3ReferenceToVideo control node that combines the Picture 1 identity reference with the single reference video depth composite (768x768, 362-frame setting). Reference audio is encoded through the LTXV audio VAE and mixed back into the final video; image presets cover 0.9MP and 0.48MP targets and the graph ships as a dual-instance configuration. 
 
This package is meant for ComfyUI and RunningHub users who want the complete swap route exactly as shipped in the episode: reference image identity, depth-composite reference video, audio preserved, and a single final export. It is the recommended starting point when evaluating the three variants of this episode. 
 
Main features: 
 
- Full-configuration character replacement with reference image identity plus reference video depth composite 
- Video Depth Anything ViT-S whole-frame depth and SeC full-clip segmentation with feathered compositing 
- MiniMax H3 hybrid FL2VA/Ref2VA INT8 base with OpenVDN DMD8 execution planning 
- Original reference audio encoded and mixed back into the final clip 
- RGB + depth person display preview output for compositing checks 
- Qwen3-VL 32B NVFP4/AWQ text encoding with MiniMax H3 video and audio VAEs 
- Dual-instance graph with 0.9MP / 0.48MP image presets and unified 24 FPS packaging 
 
Suggested workflow: 
 
Use this version as the baseline of the set: run the episode's reference image and video with the bundled prompt template, then judge whether the result already meets the goal. If facial detail or identity stability needs more work, move to the Face Refine variant with the same inputs; if audio is not required, use the Silent variant for faster iteration. 
 
For fair comparisons across the three variants of this episode, run one short test first with identical inputs, then review identity stability, motion consistency, and background drift before scaling up to full-length renders. 
 
RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
Workflow: https://www.runninghub.ai/zh-cn/post/2106048569508659202?inviteCode=rh-v1111 
 
If the results meet your expectations, you can later deploy it locally for customization. 
 
Fan Benefits: Register to get 1000 points + daily login 100 points - enjoy 4090 performance and 48 GB super power! 
 
Bilibili Updates (Mainland China &amp; Asia-Pacific) 
 
If you're in the Asia-Pacific region, you can watch the video below to see the workflow demonstration and creative breakdown. 
Bilibili Video: https://www.bilibili.com/video/BV1VjaD6jEU3/ 
 
Support Me on Ko-fi 
 
If you find my content helpful and want to support future creations, you can buy me a coffee. 
Every bit of support helps me keep creating. 
Ko-fi: https://ko-fi.com/aiksk 
 
Business Contact 
 
For collaboration or inquiries, please contact aiksk95 on WeChat. 
 
打开下方链接即可在线体验，无需安装。 
工作流：https://www.runninghub.ai/zh-cn/post/2106048569508659202?inviteCode=rh-v1111 
 
如果你觉得效果理想，也可以在本地进行自定义部署。 
 
粉丝福利：注册即可领取 1000 积分，每日登录再领 100 积分，体验 4090 和 48GB 大显存性能。 
 
B站视频（中国大陆及亚太地区） 
 
如果你在中国大陆或亚太地区，可以通过下面的视频查看工作流的实测效果与创作思路。 
B站视频：https://www.bilibili.com/video/BV1VjaD6jEU3/ 
 
我会在夸克网盘持续更新模型资源： 
https://pan.quark.cn/s/9ec095b95838

Files with SHA256: d11cb8566ceee99556c5da4a2ff16beb93ae7fae2f20f8887a33c015733b8832