<div data-youtube-video="https://youtu.be/mw3JTMK0ZmA"><iframe src="https://www.youtube.com/embed/mw3JTMK0ZmA" width="853" height="480" allowfullscreen></iframe></div><br /><br />
<p>RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2085223953907404802?inviteCode=rh-v1111</p>
<h2>MiniMax H3 + VoxCPM Digital Human - Single Character</h2>
<p>Purpose: a single-character digital-human graph. You supply one character image and one VoxCPM reference audio, write the line to be spoken, and VoxCPM generates the speech while MiniMax H3 generates the matching video performance. It is the single-subject variant of this digital-human pair.</p>
<h2>Graph (as saved)</h2>
<p>25 nodes / 31 links / 5 groups.</p>
<ul>
<li>Core: <code>MiniMaxH3ReferenceToVideo</code> saved at 1344x768, length 124, <code>ref_image_size=match</code>; <code>VRGDG_MiniMaxH3AudioDrive</code> is set to lock the original VoxCPM audio and let H3 generate picture only.</li>
<li>Models: UNETLoader <code>minimax_h3_ref2va_int8_convrot.safetensors</code>; CLIPLoader <code>qwen3vl_32b_minimax_h3_int8_convrot.safetensors</code> (type minimax); VAELoader video <code>minimax_h3_video_vae_fp16.safetensors</code> and audio <code>minimax_h3_audio_vae_fp32.safetensors</code>.</li>
<li>Sampling: <code>SamplerCustomAdvanced</code> + <code>KSamplerSelect</code> <code>res_multistep</code> + <code>BasicScheduler</code> <code>simple</code>, 20 steps, denoise 1.</li>
<li>Inputs: one LoadImage character reference, one LoadAudio VoxCPM reference audio, <code>AudioCrop</code> 0:00-0:10; ResolutionSelector 16:9 at 0.9 MP with 32 multiple; duration control 8 s.</li>
<li>Output: CreateVideo 24 fps / 8-bit, SaveVideo prefix <code>video/MiniMax_H3_VoxCPM_单人版</code>.</li>
</ul>
<h2>How to use</h2>
<p>Load the character image and the reference audio, then write the spoken line in the VoxCPM node. The note in the graph is explicit that all inputs live on the left side and that VoxCPM produces the speech before H3 drives the performance. Keep the audio locked to the generated speech if you want a clean lip-drive; switch that node if you want to drive from an external track instead.</p>
<h2>Limits and not-verified items</h2>
<p>No inference was run for this listing: no speed, VRAM, lip-sync or voice-cloning claim is made. The saved prompt, image and audio are the author's demo content, not evidence of an output. VoxCPM and the H3 checkpoints are not bundled; check their licenses separately. The cover image is a promotional still from the source video, not a frame produced by this graph.</p>
<p>Tutorial: <a href="https://www.youtube.com/watch?v=mw3JTMK0ZmA" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1kUuW6FEVs" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, audio, likeness and cover rights separately. No blanket commercial or redistribution license is asserted.</p>

MiniMax H3 + VoxCPM Digital Human - Single Character

<div data-youtube-video="https://youtu.be/mw3JTMK0ZmA"></div><br /><br />
<p>RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2085223953907404802?inviteCode=rh-v1111</p>
MiniMax H3 + VoxCPM Digital Human - Single Character
<p>Purpose: a single-character digital-human graph. You supply one character image and one VoxCPM reference audio, write the line to be spoken, and VoxCPM generates the speech while MiniMax H3 generates the matching video performance. It is the single-subject variant of this digital-human pair.</p>
Graph (as saved)
<p>25 nodes / 31 links / 5 groups.</p>
<ul>
<li>Core: <code>MiniMaxH3ReferenceToVideo</code> saved at 1344x768, length 124, <code>ref_image_size=match</code>; <code>VRGDG_MiniMaxH3AudioDrive</code> is set to lock the original VoxCPM audio and let H3 generate picture only.</li>
<li>Models: UNETLoader <code>minimax_h3_ref2va_int8_convrot.safetensors</code>; CLIPLoader <code>qwen3vl_32b_minimax_h3_int8_convrot.safetensors</code> (type minimax); VAELoader video <code>minimax_h3_video_vae_fp16.safetensors</code> and audio <code>minimax_h3_audio_vae_fp32.safetensors</code>.</li>
<li>Sampling: <code>SamplerCustomAdvanced</code> + <code>KSamplerSelect</code> <code>res_multistep</code> + <code>BasicScheduler</code> <code>simple</code>, 20 steps, denoise 1.</li>
<li>Inputs: one LoadImage character reference, one LoadAudio VoxCPM reference audio, <code>AudioCrop</code> 0:00-0:10; ResolutionSelector 16:9 at 0.9 MP with 32 multiple; duration control 8 s.</li>
<li>Output: CreateVideo 24 fps / 8-bit, SaveVideo prefix <code>video/MiniMax_H3_VoxCPM_单人版</code>.</li>
</ul>
How to use
<p>Load the character image and the reference audio, then write the spoken line in the VoxCPM node. The note in the graph is explicit that all inputs live on the left side and that VoxCPM produces the speech before H3 drives the performance. Keep the audio locked to the generated speech if you want a clean lip-drive; switch that node if you want to drive from an external track instead.</p>
Limits and not-verified items
<p>No inference was run for this listing: no speed, VRAM, lip-sync or voice-cloning claim is made. The saved prompt, image and audio are the author's demo content, not evidence of an output. VoxCPM and the H3 checkpoints are not bundled; check their licenses separately. The cover image is a promotional still from the source video, not a frame produced by this graph.</p>
<p>Tutorial: <a href="https://www.youtube.com/watch?v=mw3JTMK0ZmA" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1kUuW6FEVs" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, audio, likeness and cover rights separately. No blanket commercial or redistribution license is asserted.</p>

Files with SHA256: 610b460fcf38578855034c08bd1275a1ee2b4e44127ec6e6d5064d1cd049fefe