This three-pass MiniMax H3 Ref2VA workflow is the most intensive multi-reference redraw option in this set. It combines identity and scene references with repeated generative refinement, allowing the video to move through two enlargement stages before the final 2K-oriented reconstruction.
The active route uses the MiniMax H3 H25 hybrid loader with FL2VA and Ref2VA BF16 models, Qwen3-VL 32B MiniMax text encoding, MiniMax H3 video and audio VAEs, and connected Turbo LoRAs. The first pass uses an eight-step beta schedule. Passes two and three use short simple schedules with Euler sampling and stronger detail sigma settings. Two RTX enlargement stages provide the larger working frames for the later conditioning routes.
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 Ref2VA generative redraw
- 2K-oriented multi-reference refinement
- H25 hybrid FL2VA and Ref2VA loader
- Multiple character, prop, and scene references
- Eight-step motion-establishing first pass
- Two four-step refinement passes
- Two connected RTX enlargement stages
- Final video and audio reconstruction
Suggested workflow:
Begin with a short sequence and a small reference set. Verify that each subject remains identifiable after pass one before enabling both enlargement stages. Preserve the same subject labels and prompt structure in every pass. Use later passes for texture, facial detail, and edge cleanup; do not introduce new characters or major camera changes after the first motion pass.
RunningHub Workflow
Try the workflow online right now - no installation required.
Workflow: https://www.runninghub.ai/post/2088625751147364353?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/2088625751147364353?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 Ref2VA workflow is the most intensive multi-reference redraw option in this set. It combines identity and scene references with repeated generative refinement, allowing the video to move through two enlargement stages before the final 2K-oriented reconstruction.
The active route uses the MiniMax H3 H25 hybrid loader with FL2VA and Ref2VA BF16 models, Qwen3-VL 32B MiniMax text encoding, MiniMax H3 video and audio VAEs, and connected Turbo LoRAs. The first pass uses an eight-step beta schedule. Passes two and three use short simple schedules with Euler sampling and stronger detail sigma settings. Two RTX enlargement stages provide the larger working frames for the later conditioning routes.
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 Ref2VA generative redraw
- 2K-oriented multi-reference refinement
- H25 hybrid FL2VA and Ref2VA loader
- Multiple character, prop, and scene references
- Eight-step motion-establishing first pass
- Two four-step refinement passes
- Two connected RTX enlargement stages
- Final video and audio reconstruction
Suggested workflow:
Begin with a short sequence and a small reference set. Verify that each subject remains identifiable after pass one before enabling both enlargement stages. Preserve the same subject labels and prompt structure in every pass. Use later passes for texture, facial detail, and edge cleanup; do not introduce new characters or major camera changes after the first motion pass.
RunningHub Workflow
Try the workflow online right now - no installation required.
Workflow: https://www.runninghub.ai/post/2088625751147364353?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/2088625751147364353?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