<div data-youtube-video><iframe src="https://www.youtube.com/embed/H5L5sO7dRu8" width="640" height="360" allowfullscreen="true"></iframe></div>
https://youtu.be/H5L5sO7dRu8 
 
Creator: AI-KSK · Collection: Video 327 / six-workflow MiniMax H3 RefMod episode 
 
Workflow file: <code>01_人物RefMod_单图与文件夹批量.json</code> · Cover file: <code>人物参考工作流-1.png</code> 
 
<h3>Main features:</h3> 
 
A recurring actor is easiest to manage when all later shots point to one named reference asset rather than an ad hoc selection of input photographs. This ComfyUI graph encodes the supplied character images into an H3 RefMod that later generators can load. It is pre-encoding a reference, not training a LoRA or a new model. In the episode, the sample character is Rin, described as an adult woman with a curled dark-brown updo, pearl earrings and an ivory satin blouse. 
 
The actual node chain is <code>MiniMaxH3RefModFolderLoader</code> → <code>MiniMaxH3RefModExtract</code> → <code>MiniMaxH3RefModInspect</code> → <code>PreviewImage</code>, with an H3 video VAE supplying the encoding path. The extractor's direct image/video inputs exist but are not connected in this supplied file. That distinction matters: the image-by-image option described in the video requires you to connect your own image loader. 
 
<h3>Exact values from this JSON</h3> 
 
- Folder input: <code>Rin_Identity</code>, <code>max_items=32</code>, <code>max_frames=240</code>, <code>max_edge=1024</code>. 
- Extractor: <code>name=Rin_Identity</code>, <code>mode=Full Reference</code>, <code>concept_type=identity</code>, <code>ref_resolution=768</code>. 
- Reference settings: <code>pool_h=16</code>, <code>pool_w=16</code>, <code>latent_frames=16</code>, <code>motion_only=false</code>, <code>multiplier=1</code>, <code>save=true</code>, <code>subfolder=character</code>. 
- Selected VAE: <code>minimax_h3_video_vae_int8_convrot.safetensors</code> (must exist in your installation; not bundled). 
 
<h3>Suggested workflow:</h3> 
 
Create a folder of your own authorized character photographs under ComfyUI's <code>input</code> directory, choose that folder in the loader, and update the extractor's character description if it is not Rin. Execute and inspect the saved asset in <code>models/refmods/character/</code>. It can later be selected as <code>character/Rin_Identity</code> in the accompanying render graphs, or under whatever name you save for your own character. The original Rin images are not included with this release. 
 
<h3>Limits</h3> 
 
A saved reference can help with repeated character appearance, but it is not a learned actor checkpoint or guaranteed identity lock. Conflicting photos, missing viewpoints and off-character prompts can still change the result. The cover is promotional art showing a woman and image-reference panels, not proof of generated footage. 
 
<h3>Dependencies, tutorial and deployment</h3> 
 
This project uses ComfyUI, MiniMax H3 weights and the <a href="https://github.com/Luisacaotica/ComfyUI-MiniMaxH3Mod" rel="ugc">upstream RefMod node pack</a>; workflows involving <code>VHS_</code> nodes require <a href="https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite" rel="ugc">Video Helper Suite</a>. See <a href="https://huggingface.co/Comfy-Org/MiniMax-H3" rel="ugc">Comfy-Org's H3 model inventory</a> for official distribution guidance. The exact int8-convrot video VAE and Turbo 4-step LoRA named in the creator's JSON need separate availability/provenance checks. We distribute this workflow JSON and its original cover, not external checkpoints or third-party media. Respect applicable rights and licenses. 
 
<a href="https://www.compshare.cn/images/323moGjpvfeM?referral_code=A7fTQUYY9lCDFJwA6MMvoq" rel="ugc">Deploy AI-KSK's H3｜RefMod免训练超级参考 image on 优云智算</a>. Referral disclosure: the deployment link includes an AI-KSK referral code and may support the creator. Check the cloud provider's actual billing and environment before deploying. 
 
<a href="https://youtu.be/H5L5sO7dRu8" rel="ugc">Watch the complete six-workflow YouTube tutorial</a>. This project is one module in that episode. Description and settings were prepared by inspecting the source JSON and transcript, not from a new measured benchmark. 
 
Bilibili tutorial: https://www.bilibili.com/video/BV1F2nZ6SEA7/

MiniMax H3 RefMod Character Reference

<div data-youtube-video></div>
https://youtu.be/H5L5sO7dRu8 
 
Creator: AI-KSK · Collection: Video 327 / six-workflow MiniMax H3 RefMod episode 
 
Workflow file: <code>01_人物RefMod_单图与文件夹批量.json</code> · Cover file: <code>人物参考工作流-1.png</code> 
 
Main features: 
 
A recurring actor is easiest to manage when all later shots point to one named reference asset rather than an ad hoc selection of input photographs. This ComfyUI graph encodes the supplied character images into an H3 RefMod that later generators can load. It is pre-encoding a reference, not training a LoRA or a new model. In the episode, the sample character is Rin, described as an adult woman with a curled dark-brown updo, pearl earrings and an ivory satin blouse. 
 
The actual node chain is <code>MiniMaxH3RefModFolderLoader</code> → <code>MiniMaxH3RefModExtract</code> → <code>MiniMaxH3RefModInspect</code> → <code>PreviewImage</code>, with an H3 video VAE supplying the encoding path. The extractor's direct image/video inputs exist but are not connected in this supplied file. That distinction matters: the image-by-image option described in the video requires you to connect your own image loader. 
 
Exact values from this JSON 
 
- Folder input: <code>Rin_Identity</code>, <code>max_items=32</code>, <code>max_frames=240</code>, <code>max_edge=1024</code>. 
- Extractor: <code>name=Rin_Identity</code>, <code>mode=Full Reference</code>, <code>concept_type=identity</code>, <code>ref_resolution=768</code>. 
- Reference settings: <code>pool_h=16</code>, <code>pool_w=16</code>, <code>latent_frames=16</code>, <code>motion_only=false</code>, <code>multiplier=1</code>, <code>save=true</code>, <code>subfolder=character</code>. 
- Selected VAE: <code>minimax_h3_video_vae_int8_convrot.safetensors</code> (must exist in your installation; not bundled). 
 
Suggested workflow: 
 
Create a folder of your own authorized character photographs under ComfyUI's <code>input</code> directory, choose that folder in the loader, and update the extractor's character description if it is not Rin. Execute and inspect the saved asset in <code>models/refmods/character/</code>. It can later be selected as <code>character/Rin_Identity</code> in the accompanying render graphs, or under whatever name you save for your own character. The original Rin images are not included with this release. 
 
Limits 
 
A saved reference can help with repeated character appearance, but it is not a learned actor checkpoint or guaranteed identity lock. Conflicting photos, missing viewpoints and off-character prompts can still change the result. The cover is promotional art showing a woman and image-reference panels, not proof of generated footage. 
 
Dependencies, tutorial and deployment 
 
This project uses ComfyUI, MiniMax H3 weights and the <a href="https://github.com/Luisacaotica/ComfyUI-MiniMaxH3Mod" rel="ugc">upstream RefMod node pack</a>; workflows involving <code>VHS_</code> nodes require <a href="https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite" rel="ugc">Video Helper Suite</a>. See <a href="https://huggingface.co/Comfy-Org/MiniMax-H3" rel="ugc">Comfy-Org's H3 model inventory</a> for official distribution guidance. The exact int8-convrot video VAE and Turbo 4-step LoRA named in the creator's JSON need separate availability/provenance checks. We distribute this workflow JSON and its original cover, not external checkpoints or third-party media. Respect applicable rights and licenses. 
 
<a href="https://www.compshare.cn/images/323moGjpvfeM?referral_code=A7fTQUYY9lCDFJwA6MMvoq" rel="ugc">Deploy AI-KSK's H3｜RefMod免训练超级参考 image on 优云智算</a>. Referral disclosure: the deployment link includes an AI-KSK referral code and may support the creator. Check the cloud provider's actual billing and environment before deploying. 
 
<a href="https://youtu.be/H5L5sO7dRu8" rel="ugc">Watch the complete six-workflow YouTube tutorial</a>. This project is one module in that episode. Description and settings were prepared by inspecting the source JSON and transcript, not from a new measured benchmark. 
 
Bilibili tutorial: https://www.bilibili.com/video/BV1F2nZ6SEA7/

Files with SHA256: 9bdf99b8cb1b707374a53261da19470611ba0d597597843b835a3cee6064ca26