我不小心把一个用于伦敦犯罪黑色电影的 Seedance 提示词粘贴进了 Krea 2。

I accidentally pasted one of my Seedance prompts for London crime noir into Krea 2. Instead of just giving me a single image, it came back with something more like a shot burst contact sheet. What surprised me was how well it held onto the visual rules I’d built into the prompt i.e. colour palette, framing, subject placement, symmetry negative space, overall moodboard vibe This is something I struggle to maintain when trying to expand Midjourney images into more scenes or a larger world with the same consistent aesthetic. Because it’s generating a shot burst, a lot of the individual frames are imperfect or partially broken but you isolate the frames you like, pull compositions that feel strong, find interesting bits of framing or atmosphere, then recreate or upscale from there. I'm thinking this is less about getting perfect finished images and more about visualising the cinematic language of the scenes/world I'm trying to build. Try it for yourself. Put a video gen prompt into Krea 2 and see what you get. (The below was Krea 2 Large with a custom moodboard)

我不小心把一个用于伦敦犯罪黑色电影的 Seedance 提示词粘贴进了 Krea 2。 它没有只给我返回一张单一的图像，而是给出了更像是「shot burst」（连拍）那样的 contact sheet（联系表）。 让我感到惊讶的是，它很好地保留了我在提示词里构建的视觉规则，也就是：色彩搭配、构图、主体放置、对称负空间、整体的 moodboard 氛围。 当我试图把 Midjourney 的图像扩展成更多场景，或者扩展成一个具有一致美学的更大世界时，这是我很难维持住的东西。 因为它生成的是一组连拍，其中很多单独的帧是不完美的，或者部分崩坏，但你可以挑出你喜欢的帧，抽取感觉很强的构图，找到有趣的构图或氛围片段，然后基于它们重新创作或放大。 我觉得这与其说是为了得到完美的成品图像，不如说是为了把我想要搭建的场景/世界的电影化语言可视化出来。 你自己试试看。把一个视频生成的提示词放进 Krea 2，看看你能得到什么。 （下面这张是 Krea 2 Large 配合自定义 moodboard 生成的）