This three-pass MiniMax H3 FL2VA workflow is designed for aggressive generative video enlargement toward a 2K-class result. Instead of relying on a single upscale, it alternates MiniMax H3 generation and spatial enlargement so that each later pass can reconstruct detail at the actual larger frame size.
The workflow uses MiniMax H3 FL2VA BF16, Qwen3-VL 32B MiniMax encoding, MiniMax H3 video and audio VAEs, and connected four-step and eight-step Turbo LoRAs. Pass one establishes the video with an eight-step beta schedule. RTX enlargement stages sit between the generation passes. Passes two and three rebuild conditioning at the measured output dimensions and use shorter simple schedules for controlled refinement rather than restarting the motion from scratch.
The workflow is packaged as a ready-to-run ComfyUI graph for users who want to compare generative enlargement against conventional video scaling. Keep the original source, prompt, and duration unchanged when comparing versions so that visible differences come from the workflow structure rather than from changed creative inputs.
Main features:
- Three-stage FL2VA generative enlargement
- 2K-oriented final refinement route
- First-frame and last-frame compatible conditioning
- Eight-step first pass for motion establishment
- Short detail passes after each enlargement
- Two connected RTX enlargement stages
- Actual frame-size measurement before regeneration
- MiniMax H3 video and audio decoding
Suggested workflow:
Complete a stable first pass before enabling the full three-stage route. Use the same identity and scene description across all passes. Let later prompts emphasize clarity and material detail instead of changing composition. For portrait sources, verify the selected aspect ratio and test a short duration first, because every additional pass increases both processing time and the chance of small identity changes.
RunningHub Workflow
Try the workflow online right now - no installation required.
Workflow: https://www.runninghub.ai/post/2089197182574870530?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 & Asia-Pacific)
Watch the workflow demonstration and creative breakdown here:
Bilibili Video: https://www.bilibili.com/video/BV1kZbQ6xEMg/
Support Me on Ko-fi
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/2089197182574870530?inviteCode=rh-v1111
如果觉得效果理想，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 和 48 GB 大显存性能。
B站视频：https://www.bilibili.com/video/BV1kZbQ6xEMg/
我会在夸克网盘持续更新模型资源：
https://pan.quark.cn/s/07bdc81784ce
Description
This three-pass MiniMax H3 FL2VA workflow is designed for aggressive generative video enlargement toward a 2K-class result. Instead of relying on a single upscale, it alternates MiniMax H3 generation and spatial enlargement so that each later pass can reconstruct detail at the actual larger frame size.
The workflow uses MiniMax H3 FL2VA BF16, Qwen3-VL 32B MiniMax encoding, MiniMax H3 video and audio VAEs, and connected four-step and eight-step Turbo LoRAs. Pass one establishes the video with an eight-step beta schedule. RTX enlargement stages sit between the generation passes. Passes two and three rebuild conditioning at the measured output dimensions and use shorter simple schedules for controlled refinement rather than restarting the motion from scratch.
The workflow is packaged as a ready-to-run ComfyUI graph for users who want to compare generative enlargement against conventional video scaling. Keep the original source, prompt, and duration unchanged when comparing versions so that visible differences come from the workflow structure rather than from changed creative inputs.
Main features:
- Three-stage FL2VA generative enlargement
- 2K-oriented final refinement route
- First-frame and last-frame compatible conditioning
- Eight-step first pass for motion establishment
- Short detail passes after each enlargement
- Two connected RTX enlargement stages
- Actual frame-size measurement before regeneration
- MiniMax H3 video and audio decoding
Suggested workflow:
Complete a stable first pass before enabling the full three-stage route. Use the same identity and scene description across all passes. Let later prompts emphasize clarity and material detail instead of changing composition. For portrait sources, verify the selected aspect ratio and test a short duration first, because every additional pass increases both processing time and the chance of small identity changes.
RunningHub Workflow
Try the workflow online right now - no installation required.
Workflow: https://www.runninghub.ai/post/2089197182574870530?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 & Asia-Pacific)
Watch the workflow demonstration and creative breakdown here:
Bilibili Video: https://www.bilibili.com/video/BV1kZbQ6xEMg/
Support Me on Ko-fi
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/2089197182574870530?inviteCode=rh-v1111
如果觉得效果理想，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 和 48 GB 大显存性能。
B站视频：https://www.bilibili.com/video/BV1kZbQ6xEMg/
我会在夸克网盘持续更新模型资源：
https://pan.quark.cn/s/07bdc81784ce