MiniMax H3 I2V Test: Precise Cinematic Control via Structured Prompts

crazeum · reddit · 2026-08-05

User crazeum shared an image-to-video test using the MiniMax H3 model.

The test utilizes highly structured prompts to dictate the video generation, precisely defining timestamps, camera movements (e.g., slow push-in, static tracking), and character actions.

The author noted that audio generation can be hit-or-miss (such as erratic audio at the beginning or inconsistent Picard voice), suggesting that users might need Ref2Vid nodes or further prompt refinement for better control.

Original post →

More from Multimodal

Multimodal channel →