<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/2084850865038520322?inviteCode=rh-v1111
<h2>MiniMax H3 First-and-last-frame video</h2>
A MiniMax H3 omni-modal video graph exported from episode 288. 27 nodes / 32 links. Two LoadImage inputs: the first image is the starting frame, the second the ending frame. 16:9 at 0.8 MP, 10 s.
<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>
Prompt contract is positional: preserve the same subject across both frames. The saved example is an anime-style basketball player.
<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 First-and-Last-Frame Video

<div data-youtube-video="https://youtu.be/RhszI-72alU"></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2084850865038520322?inviteCode=rh-v1111
MiniMax H3 First-and-last-frame video
A MiniMax H3 omni-modal video graph exported from episode 288. 27 nodes / 32 links. Two LoadImage inputs: the first image is the starting frame, the second the ending frame. 16:9 at 0.8 MP, 10 s.
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
Prompt contract is positional: preserve the same subject across both frames. The saved example is an anime-style basketball player.
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.

minimaxH3FirstAndLast_v10.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> 
 
This workflow uses the MiniMax H3 reference-to-video route for a controlled transition between two visual states. It is suited to shots where both the opening composition and the final reveal need to remain intentional instead of drifting through an unconstrained generation. 
 
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 video VAE and Qwen3-VL text encoder 
- Start and end image guidance for a structured visual transition 
- Prompt control for camera movement, subject action, scene continuity, and audio 
- Connected 9:16 portrait widescreen resolution route at the 0.8 megapixel setting 
- 20-step simple scheduler with res_multistep sampling 
- Frame duration conversion for the H3 temporal block structure 
- Two SaveVideo output paths are present; use the connected final route for publication 
 
Suggested workflow: 
 
Prepare the start and end images with compatible subject identity and composition. Describe the transition as a sequence, including what moves first and what must be visible at the final moment. Use the portrait route for vertical content, then adjust duration and prompt timing before changing the core sampler. 
 
⚡ RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
👉 Workflow: https://www.runninghub.ai/post/2084850865038520322?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/2084850865038520322?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 Start-End Frame Workflow

<div data-youtube-video="https://youtu.be/RhszI-72alU"></div> 
 
This workflow uses the MiniMax H3 reference-to-video route for a controlled transition between two visual states. It is suited to shots where both the opening composition and the final reveal need to remain intentional instead of drifting through an unconstrained generation. 
 
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 video VAE and Qwen3-VL text encoder 
- Start and end image guidance for a structured visual transition 
- Prompt control for camera movement, subject action, scene continuity, and audio 
- Connected 9:16 portrait widescreen resolution route at the 0.8 megapixel setting 
- 20-step simple scheduler with res_multistep sampling 
- Frame duration conversion for the H3 temporal block structure 
- Two SaveVideo output paths are present; use the connected final route for publication 
 
Suggested workflow: 
 
Prepare the start and end images with compatible subject identity and composition. Describe the transition as a sequence, including what moves first and what must be visible at the final moment. Use the portrait route for vertical content, then adjust duration and prompt timing before changing the core sampler. 
 
⚡ RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
👉 Workflow: https://www.runninghub.ai/post/2084850865038520322?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/2084850865038520322?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

Files with SHA256: 25448b7d6820f2384cf1ffc69c875be64d6b313ea37a3749b21527c500d95e7c