gpt-imageVIDEO

Generating character and scene assets with GPT Image 2.5 before making video

The author describes an AI video workflow: first use GPT Image 2.5 to generate character, scene and prop reference images for consistency, then hand them to Muse to produce the video.

Jarvis｜AI 商业实验室

@jarvis11xArchived 26 days agoOriginally posted 2026.10.05

Jarvis｜AI 商业实验室 @jarvis11xOriginally posted 2026.10.05

Open original on X

用 Muse 生成视频我都是先提前生成人物、场景、道具锁定资产素材，这样才能确保人物的镜头的一致性一开始人物素材我都是让 GPT image 2.5 生成的现在我直接让Muse出图了，Muse image能力其实也不差生成视频前，只要加一句提示词：所有分镜生成前，请先创作人物、场景、道具的参考图作为创作资产包，并在对应的分镜生成时加上参考图

This post quotes another post: x.com/jarvis11x/status/2104397648438476868

x.com/jarvis11x/status/2106960040246587835Archived 2026.10.06 10:09