Developer Creates Custom GPT Assistant for MiniMax H3 Video Prompts

midnitefox · reddit · 2026-08-05

To simplify prompt engineering for the MiniMax H3 video model, a developer created a dedicated Prompting Assistant GPT. Based on the official MiniMax guide, it supports four workflows: text-to-video, image-to-video (opening frame), transitions between first and final images, and ending on a specific final frame.

The assistant asks for the active workflow before generating to apply the correct structure. For text-to-video, it recommends resolution and length based on scene complexity, actions, and camera movement. It also handles advanced structured instructions like shot timelines, camera movement, prop continuity, multi-speaker dialogue formatting, and diegetic sound, helping users quickly generate professional prompts.

Original post →

More from Multimodal

Multimodal channel →