Minimax_H3_Zs05_双采_加速_Lantent_Upscaler参考生视频
UpdateSep 15, 2026 19:16PublishedSep 15, 2026 19:16
1
0
0
Avatar
Lorena
Type
Workflow
Basic Model
MiniMax H3
Published time
Sep 15, 2026 19:16
File Signature
235c06301b6d51950ffc8ccb3f194b4b4a09cb6de23b922361cdaf032c43c5a0

A reference-to-video workflow combining Z-Image and MiniMax H3 for an integrated image-to-video generation pipeline. Z-Image is used to generate and prepare high-quality visual references, while MiniMax H3 handles video generation with a dual-pass sampling strategy and Latent Upscaler for improved detail, motion quality, and final resolution. The workflow also includes acceleration optimizations to balance generation quality and efficiency. Well suited for character-driven videos, cinematic scenes, stylized animation, and other reference-based video generation tasks. Generating a 15-second video takes approximately 7 minutes and costs $0.38.

Gallery

No creation yet