This two-pass MiniMax H3 FL2VA workflow is built for high-resolution video generation guided by a first frame and an optional last frame. The first pass establishes motion and composition, while the second pass redraws the enlarged result for improved texture, edge definition, and frame-to-frame consistency.
The active model route uses MiniMax H3 FL2VA BF16 with Qwen3-VL 32B MiniMax encoding, MiniMax H3 video and audio VAEs, and both four-step and eight-step FL2V Turbo LoRAs. Pass one uses an eight-step beta schedule with the dynamic sigma-shift route. The intermediate result is spatially enlarged, measured at its actual dimensions, and rebuilt in pass two with a short simple schedule and the detail-oriented sigma settings.
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:
- Two-stage FL2VA generative upscale structure
- First-frame and optional last-frame guidance
- MiniMax H3 FL2VA BF16 model
- Four-step and eight-step Turbo LoRA routes
- Eight-step beta schedule for the first pass
- Intermediate RTX spatial enlargement
- Second-pass conditioning rebuilt at the real enlarged size
- Separate video and audio VAE output paths
Suggested workflow:
Use a clean first frame and add a last frame only when the ending composition must be constrained. Test the motion pass before increasing the scale. For pass two, keep the prompt focused on materials, facial detail, clothing, and background clarity rather than introducing new actions. If the last-frame transition becomes unstable, reduce scene complexity or remove conflicting references.
RunningHub Workflow
Try the workflow online right now - no installation required.
Workflow: https://www.runninghub.ai/post/2088625742452572161?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/2088625742452572161?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 two-pass MiniMax H3 FL2VA workflow is built for high-resolution video generation guided by a first frame and an optional last frame. The first pass establishes motion and composition, while the second pass redraws the enlarged result for improved texture, edge definition, and frame-to-frame consistency.
The active model route uses MiniMax H3 FL2VA BF16 with Qwen3-VL 32B MiniMax encoding, MiniMax H3 video and audio VAEs, and both four-step and eight-step FL2V Turbo LoRAs. Pass one uses an eight-step beta schedule with the dynamic sigma-shift route. The intermediate result is spatially enlarged, measured at its actual dimensions, and rebuilt in pass two with a short simple schedule and the detail-oriented sigma settings.
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:
- Two-stage FL2VA generative upscale structure
- First-frame and optional last-frame guidance
- MiniMax H3 FL2VA BF16 model
- Four-step and eight-step Turbo LoRA routes
- Eight-step beta schedule for the first pass
- Intermediate RTX spatial enlargement
- Second-pass conditioning rebuilt at the real enlarged size
- Separate video and audio VAE output paths
Suggested workflow:
Use a clean first frame and add a last frame only when the ending composition must be constrained. Test the motion pass before increasing the scale. For pass two, keep the prompt focused on materials, facial detail, clothing, and background clarity rather than introducing new actions. If the last-frame transition becomes unstable, reduce scene complexity or remove conflicting references.
RunningHub Workflow
Try the workflow online right now - no installation required.
Workflow: https://www.runninghub.ai/post/2088625742452572161?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/2088625742452572161?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