<div data-youtube-video="https://youtu.be/RhszI-72alU"><iframe src="https://www.youtube.com/embed/RhszI-72alU" width="853" height="480" allowfullscreen></iframe></div> 
 
This workflow is a local ComfyUI text-to-video setup for MiniMax H3. It turns one prompt into a cinematic video with synchronized visual and audio generation, making it useful for concept shots, short narrative clips, music-driven scenes, and fast model testing. 
 
The workflow uses the MiniMax H3 local ComfyUI graph with the connected model, VAE, encoder, sampler, resolution, and video output chain shown in the JSON. 
 
Main features: 
 
- MiniMax H3 FL2VA diffusion model with the connected video VAE and Qwen3-VL text encoder 
- Text-driven video generation with prompt control over shots, camera movement, dialogue, sound effects, and music 
- Resolution Selector with a connected 16:9 widescreen route at the 0.8 megapixel setting 
- 20-step simple scheduler with the res_multistep sampler route 
- Duration input converted to a valid video frame length for the H3 frame-block requirement 
- CreateVideo and SaveVideo nodes for a direct video output path 
- Native audio-aware generation instead of adding a separate soundtrack afterward 
 
Suggested workflow: 
 
Load the MiniMax H3 model, video VAE, and Qwen3-VL encoder first. Write the visual action and audio direction in one prompt, choose the duration, then select the connected 16:9 resolution. Keep the sampler and scheduler settings stable while testing prompts; adjust duration and prompt structure before changing the model route. 
 
⚡ RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
👉 Workflow: https://www.runninghub.ai/post/2084544656423088130?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/BV1MrMk69Ejk/ 
 
☕ 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/post/2084544656423088130?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 与 48 GB 超大显存性能。 
 
📵 B站视频（中国大陆及亚洲地区） 
 
如果你在中国大陆或亚洲地区，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/BV1MrMk69Ejk/ 
 
我会在夸克网盘持续更新模型资源： 
👉 https://pan.quark.cn/s/07bdc81784ce

MiniMax H3 Text-to-Video Workflow

<div data-youtube-video="https://youtu.be/RhszI-72alU"></div> 
 
This workflow is a local ComfyUI text-to-video setup for MiniMax H3. It turns one prompt into a cinematic video with synchronized visual and audio generation, making it useful for concept shots, short narrative clips, music-driven scenes, and fast model testing. 
 
The workflow uses the MiniMax H3 local ComfyUI graph with the connected model, VAE, encoder, sampler, resolution, and video output chain shown in the JSON. 
 
Main features: 
 
- MiniMax H3 FL2VA diffusion model with the connected video VAE and Qwen3-VL text encoder 
- Text-driven video generation with prompt control over shots, camera movement, dialogue, sound effects, and music 
- Resolution Selector with a connected 16:9 widescreen route at the 0.8 megapixel setting 
- 20-step simple scheduler with the res_multistep sampler route 
- Duration input converted to a valid video frame length for the H3 frame-block requirement 
- CreateVideo and SaveVideo nodes for a direct video output path 
- Native audio-aware generation instead of adding a separate soundtrack afterward 
 
Suggested workflow: 
 
Load the MiniMax H3 model, video VAE, and Qwen3-VL encoder first. Write the visual action and audio direction in one prompt, choose the duration, then select the connected 16:9 resolution. Keep the sampler and scheduler settings stable while testing prompts; adjust duration and prompt structure before changing the model route. 
 
⚡ RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
👉 Workflow: https://www.runninghub.ai/post/2084544656423088130?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/BV1MrMk69Ejk/ 
 
☕ 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/post/2084544656423088130?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 与 48 GB 超大显存性能。 
 
📵 B站视频（中国大陆及亚洲地区） 
 
如果你在中国大陆或亚洲地区，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/BV1MrMk69Ejk/ 
 
我会在夸克网盘持续更新模型资源： 
👉 https://pan.quark.cn/s/07bdc81784ce

01_minimax-h3-text-to-video-workflow.json

v1.0

<div data-youtube-video="https://youtu.be/RhszI-72alU"><iframe src="https://www.youtube.com/embed/RhszI-72alU" width="853" height="480" allowfullscreen></iframe></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2084544656423088130?inviteCode=rh-v1111
<h2>MiniMax H3 Text-to-video</h2>
A MiniMax H3 omni-modal video graph exported from episode 288. 23 nodes / 24 links, one branch. The saved prompt is a 10-second 16:9 3840x2160 first-person MotoGP POV scene. 16:9 at 0.8 MP.
<h2>Saved configuration</h2>
<ul>
<li>UNETLoader <code>minimax_h3_fl2va_bf16.safetensors</code>; CLIPLoader <code>qwen3vl_32b_minimax_h3_int8_convrot.safetensors</code> (type minimax); VAELoader video <code>minimax_h3_video_vae_fp16.safetensors</code> + audio <code>minimax_h3_audio_vae_fp32.safetensors</code>.</li>
<li>The saved graphs contain no LoRA nodes.</li>
<li>Sampling: <code>KSamplerSelect</code> res_multistep with <code>BasicScheduler</code> simple, 20 steps, denoise 1; <code>SamplerCustomAdvanced</code> for the custom chain.</li>
<li>Conditioning: <code>MiniMaxH3ImageToVideo</code> / <code>MiniMaxH3ReferenceToVideo</code> saved at 1344x768 / 124 / match where present; ResolutionSelector and duration values are quoted per item above; output goes to SaveVideo prefix <code>video/MiniMax_H3</code>.</li>
</ul>
<h2>How to use</h2>
No image input at all: the prompt is the only content input. MiniMaxH3ImageToVideo runs as a pure T2V conditioning node.
<h2>Limits and not-verified items</h2>
No inference was run for this listing: no speed, VRAM, audio-sync or quality claim is made. Values are quoted from the saved nodes. The graph's notes link the official Comfy-Org/MiniMax-H3 model files and the MiniMax blog; those notes are preserved and this listing does not claim to be the original template author. Weights, encoder and VAEs are not bundled; check their licenses separately. The cover image is a promotional still, not a frame produced by this graph.
Tutorial: <a href="https://www.youtube.com/watch?v=RhszI-72alU" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1MrMk69Ejk" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, image/audio and cover rights separately. No blanket commercial or redistribution license is asserted.

MiniMax H3 Text-to-Video

<div data-youtube-video="https://youtu.be/RhszI-72alU"></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2084544656423088130?inviteCode=rh-v1111
MiniMax H3 Text-to-video
A MiniMax H3 omni-modal video graph exported from episode 288. 23 nodes / 24 links, one branch. The saved prompt is a 10-second 16:9 3840x2160 first-person MotoGP POV scene. 16:9 at 0.8 MP.
Saved configuration
<ul>
<li>UNETLoader <code>minimax_h3_fl2va_bf16.safetensors</code>; CLIPLoader <code>qwen3vl_32b_minimax_h3_int8_convrot.safetensors</code> (type minimax); VAELoader video <code>minimax_h3_video_vae_fp16.safetensors</code> + audio <code>minimax_h3_audio_vae_fp32.safetensors</code>.</li>
<li>The saved graphs contain no LoRA nodes.</li>
<li>Sampling: <code>KSamplerSelect</code> res_multistep with <code>BasicScheduler</code> simple, 20 steps, denoise 1; <code>SamplerCustomAdvanced</code> for the custom chain.</li>
<li>Conditioning: <code>MiniMaxH3ImageToVideo</code> / <code>MiniMaxH3ReferenceToVideo</code> saved at 1344x768 / 124 / match where present; ResolutionSelector and duration values are quoted per item above; output goes to SaveVideo prefix <code>video/MiniMax_H3</code>.</li>
</ul>
How to use
No image input at all: the prompt is the only content input. MiniMaxH3ImageToVideo runs as a pure T2V conditioning node.
Limits and not-verified items
No inference was run for this listing: no speed, VRAM, audio-sync or quality claim is made. Values are quoted from the saved nodes. The graph's notes link the official Comfy-Org/MiniMax-H3 model files and the MiniMax blog; those notes are preserved and this listing does not claim to be the original template author. Weights, encoder and VAEs are not bundled; check their licenses separately. The cover image is a promotional still, not a frame produced by this graph.
Tutorial: <a href="https://www.youtube.com/watch?v=RhszI-72alU" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV1MrMk69Ejk" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, image/audio and cover rights separately. No blanket commercial or redistribution license is asserted.

Files with SHA256: 5adc8acbf10cfd232543926d1ee3e19f02b43cd954487385dc1bcfa990e230f3