Project: Purge 清洗计划
1,4282026-03-30 16:58:56
该内容使用了AI生成技术,请谨慎甄别
浮光Flow
浮光Flow
粉丝·1作品
在未来都市,落魄的硬汉Matt Leroy坚信自己是遭顶替的人类CEO本尊。为了夺回一切,他一路杀入财阀总部,击毙了那个和他外貌一模一样的“冒牌货”。然而地上留下的鲜血,却撕开了一个绝望的真相:死去的才是真正的人类。而自己只是个被植入原主记忆的机器武器,专门被造出来执行这场弑主的“清洗计划”。

创作工具

图片
Nano-Banana 2
视频
万相视频 2.5,即梦视频3.5,可灵 3.0

分镜拆解

分镜 highlight_a1261x96r5b_17748611361820
置顶
00:48-00:49
分镜 highlight_a1261x96r5b_17748611361821
置顶
04:06-04:13
分镜 highlight_a1261x96r5b_17748611361822
置顶
04:22-04:28
分镜 highlight_a1261x96r5b_17748611361823
置顶
06:07-06:11
分镜 highlight_a1261x96r5b_17748611361824
置顶
06:34-06:37
分镜 highlight_a1261x96r5b_17748611361825
置顶
06:46-06:50
分镜 1
主要模型
Nano-Banana 2
视频提示词
{ "prompt": "(Masterpiece, best quality, photorealistic:1.4), cinematic movie still, wide shot, perfect symmetry, eye level. \n\n[Environment]: A vast, minimalist abstract space. The walls are enormous panels of glowing deep crimson red light (or translucent red glass). \n\n[Composition]: Strong horizontal lines. No windows, just an enclosed, suffocatingly large red void. \n\n[Centerpiece]: A sleek, low-profile black monolithic desk (horizontal orientation) positioned perfectly in the center. \n\n[Character]: Matt Leroy (CEO) sitting behind the desk, framed dead center. \n- He looks small in the vast room but commands total attention. \n- Wearing a sharp black suit (silhouette against the red background). \n- His expression is cold, detached, almost bored. \n- Steepled hands (fingertips touching) on the desk. \n\n[Atmosphere]: Oppressive, heavy atmosphere, danger, sci-fi authoritarianism, 'The Shining' level of uneasiness. \n\n[Lighting]: The light comes entirely from the glowing red walls, casting him in silhouette or rim light. \n\n[Style]: Roger Deakins cinematography (Blade Runner 2049 style), bold use of color, high contrast, 8k uhd.", "negative_prompt": "(bright lights, white room:1.5), (clutter, messy:1.5), (windows, city view:1.5), (warm orange, yellow:1.5), (cartoon, anime:1.3), furniture, plants, carpet, realistic office.", "aspect_ratio": "16:9", "seed": -1 }
如今AI模型的迭代已经到了极其恐怖的速度,我认为作为创作者而言,这是最好的时代。 但更加强大的模型需要更具备经验的使用者才能调配出更精准的画面“配方”,我在制作如此复杂的美术场景时,ai工具成为了我最好的美术助手。gemini能够精准地根据我的需求去生成适配于不同平台和模型的提示词,如mj v7、banana pro等。确定美术场景的氛围图,是我认为ai作品极为重要的一步,通过氛围图确定核心提示词和画面风格,再将提示词或图片直接投喂给模型去生产分镜,得到的结果往往是喜人的。此次的分镜图片产出主要采用了“banana直出+修改”的方式,但修改最好不要超过两次,否则原图的质感会被算法磨平,最好的办法依旧是依靠复杂的提示词直出。 这次我们团队使用的视频模型全部来自国内平台,可灵o3、wan2.6、viduQ3、seedance1.5pro(seedacne2.0因人脸审核问题无法使用),每个平台各有所长,o3的真实感更强,逻辑连贯性不错;wan2.6对于动态图形的理解很强,响应速度也很快;vidu打斗效果好。拥有这些模型,几乎你所有的想象都能达成,这在一年之前是无法想象的事情,不必再因技术的问题去妥协你的表达。