<div data-youtube-video="https://youtu.be/ZP3rffpcIXg"><iframe src="https://www.youtube.com/embed/ZP3rffpcIXg" width="853" height="480" allowfullscreen></iframe></div><br /><br /><p>RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2094644471542673409?inviteCode=rh-v1111</p><h2>Graph</h2><p>MiniMax H3 Ref2VA - Multi-Reference with Keyframe Story Beats. 87 nodes, 141 links, 40 node types. No LoRA nodes; sampling runs through custom sampler nodes.</p><p>No inference was run for this listing: no rendered clip, no comparison sheet, no timing.</p><h2>Files</h2><ul><li><code>MiniMax+H3｜Ref2VA+多参考+关键帧剧情控制｜Motion+Lab+动作.json</code></li><li><code>f8bc5194-bf3f-40e3-b3e4-1fd104a11cf6.png</code> (promotional still, not a graph output)</li></ul><h2>Tutorial and rights</h2><p>Source video: <a href="https://www.youtube.com/watch?v=ZP3rffpcIXg" rel="ugc">https://www.youtube.com/watch?v=ZP3rffpcIXg</a>. Chinese mirror: <a href="https://www.bilibili.com/video/BV1L4tn6WEVb" rel="ugc">https://www.bilibili.com/video/BV1L4tn6WEVb</a>. Weights and input media are not bundled; check licences separately.</p>

MiniMax H3 Ref2VA - Multi-Reference with Keyframe Story Beats

<div data-youtube-video="https://youtu.be/ZP3rffpcIXg"></div><br /><br /><p>RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2094644471542673409?inviteCode=rh-v1111</p>Graph<p>MiniMax H3 Ref2VA - Multi-Reference with Keyframe Story Beats. 87 nodes, 141 links, 40 node types. No LoRA nodes; sampling runs through custom sampler nodes.</p><p>No inference was run for this listing: no rendered clip, no comparison sheet, no timing.</p>Files<ul><li><code>MiniMax+H3｜Ref2VA+多参考+关键帧剧情控制｜Motion+Lab+动作.json</code></li><li><code>f8bc5194-bf3f-40e3-b3e4-1fd104a11cf6.png</code> (promotional still, not a graph output)</li></ul>Tutorial and rights<p>Source video: <a href="https://www.youtube.com/watch?v=ZP3rffpcIXg" rel="ugc">https://www.youtube.com/watch?v=ZP3rffpcIXg</a>. Chinese mirror: <a href="https://www.bilibili.com/video/BV1L4tn6WEVb" rel="ugc">https://www.bilibili.com/video/BV1L4tn6WEVb</a>. Weights and input media are not bundled; check licences separately.</p>

minimaxH3Ref2vaMulti_v10.json

v1.0

<div data-youtube-video="https://youtu.be/ZP3rffpcIXg"><iframe src="https://www.youtube.com/embed/ZP3rffpcIXg" width="853" height="480" allowfullscreen></iframe></div><br />
<br />
MiniMax H3 Ref2VA multi-reference + keyframe story control, with Motion Lab action repair - a single ComfyUI workflow that turns multiple reference pictures into one coherent cinematic video.<br />
<br />
This workflow is built around the MiniMax H3 Ref2VA model (minimax_h3_ref2va_int8_convrot.safetensors) and drives it through a multi-pass SamplerCustomAdvanced pipeline with res_multistep sampling, a Qwen3-VL 32B text encoder (qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors), separate video VAE (minimax_h3_video_vae_fp16) and audio VAE (minimax_h3_audio_vae_fp32), plus a compatible 3D latent upscaler pass at 1.5x before the final save.<br />
<br />
Main features:<br />
- Multi-reference fusion: bind subject, outfit, environment and prop to &lt;Picture N&gt; slots in the prompt so identity stays consistent across shots.<br />
- Keyframe story control: anchor specific frames to narrative beats inside one continuous generation.<br />
- Motion Lab action repair: fixes motion artifacts on fast-action sequences.<br />
- Two-stage output: an 8-second PASS1 preview plus the final AllInOne video, with a 3D HiRes 1.5x upscale pass before saving.<br />
<br />
Suggested workflow:<br />
1. Load your reference pictures (subject / outfit / environment / prop) into the image slots.<br />
2. Write subject_definitions for each &lt;Picture N&gt;, then one summary line describing the full sequence.<br />
3. Run PASS1 to check composition, then run the final pass with the 1.5x 3D upscale enabled.<br />
<br />
Notes: 5 output-reachable nodes are intentionally bypassed in this build; their effects are disabled by design. Requires MiniMax H3 Ref2VA int8 weights, the Qwen3-VL 32B encoder and both video/audio VAEs.<br />
<br />
RunningHub (one-click run): https://www.runninghub.ai/zh-cn/post/2095021906982813697?inviteCode=rh-v1111<br />
Bilibili tutorial: https://www.bilibili.com/video/BV1L4tn6WEVb/<br />
<br />
Models &amp; resources:<br />
MiniMaxAI/MiniMax-H3-Turbo-Lora: https://huggingface.co/MiniMaxAI/MiniMax-H3-Turbo-Lora<br />
lightx2v/Minimax-h3-Turbo: https://huggingface.co/lightx2v/Minimax-h3-Turbo<br />
fal/MiniMax-H3-Realism-People-LoRA: https://huggingface.co/fal/MiniMax-H3-Realism-People-LoRA<br />
Model pack (Quark): https://pan.quark.cn/s/9ec095b95838<br />
<br />
RunningHub signup with invite code rh-v1111 gets 1000 RH coins: https://www.runninghub.ai/?inviteCode=rh-v1111

MiniMax H3 Ref2VA Multi-Reference Story Control

<div data-youtube-video="https://youtu.be/ZP3rffpcIXg"></div><br />
<br />
MiniMax H3 Ref2VA multi-reference + keyframe story control, with Motion Lab action repair - a single ComfyUI workflow that turns multiple reference pictures into one coherent cinematic video.<br />
<br />
This workflow is built around the MiniMax H3 Ref2VA model (minimax_h3_ref2va_int8_convrot.safetensors) and drives it through a multi-pass SamplerCustomAdvanced pipeline with res_multistep sampling, a Qwen3-VL 32B text encoder (qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors), separate video VAE (minimax_h3_video_vae_fp16) and audio VAE (minimax_h3_audio_vae_fp32), plus a compatible 3D latent upscaler pass at 1.5x before the final save.<br />
<br />
Main features:<br />
- Multi-reference fusion: bind subject, outfit, environment and prop to &lt;Picture N&gt; slots in the prompt so identity stays consistent across shots.<br />
- Keyframe story control: anchor specific frames to narrative beats inside one continuous generation.<br />
- Motion Lab action repair: fixes motion artifacts on fast-action sequences.<br />
- Two-stage output: an 8-second PASS1 preview plus the final AllInOne video, with a 3D HiRes 1.5x upscale pass before saving.<br />
<br />
Suggested workflow:<br />
1. Load your reference pictures (subject / outfit / environment / prop) into the image slots.<br />
2. Write subject_definitions for each &lt;Picture N&gt;, then one summary line describing the full sequence.<br />
3. Run PASS1 to check composition, then run the final pass with the 1.5x 3D upscale enabled.<br />
<br />
Notes: 5 output-reachable nodes are intentionally bypassed in this build; their effects are disabled by design. Requires MiniMax H3 Ref2VA int8 weights, the Qwen3-VL 32B encoder and both video/audio VAEs.<br />
<br />
RunningHub (one-click run): https://www.runninghub.ai/zh-cn/post/2095021906982813697?inviteCode=rh-v1111<br />
Bilibili tutorial: https://www.bilibili.com/video/BV1L4tn6WEVb/<br />
<br />
Models &amp; resources:<br />
MiniMaxAI/MiniMax-H3-Turbo-Lora: https://huggingface.co/MiniMaxAI/MiniMax-H3-Turbo-Lora<br />
lightx2v/Minimax-h3-Turbo: https://huggingface.co/lightx2v/Minimax-h3-Turbo<br />
fal/MiniMax-H3-Realism-People-LoRA: https://huggingface.co/fal/MiniMax-H3-Realism-People-LoRA<br />
Model pack (Quark): https://pan.quark.cn/s/9ec095b95838<br />
<br />
RunningHub signup with invite code rh-v1111 gets 1000 RH coins: https://www.runninghub.ai/?inviteCode=rh-v1111

Files with SHA256: 5be1d49fe54a0b05665928373ada4cc305c3b1a425107ab349a86b08268be5b6