使用 Grok Imagine Video 1.5 制作(参考生成视频,3 个图像槽位)。, 图片 1
1/4
使用 Grok Imagine Video 1.5 制作(参考生成视频,3 个图像槽位)。, 图片 2
2/4
使用 Grok Imagine Video 1.5 制作(参考生成视频,3 个图像槽位)。, 图片 3
3/4
使用 Grok Imagine Video 1.5 制作(参考生成视频,3 个图像槽位)。, 图片 4
4/4

PROMPT RECORD · 公开作品详情

使用 Grok Imagine Video 1.5 制作(参考生成视频,3 个图像槽位)。

公开 AI 图像创作记录,包含原图、提示词、使用模型、作者、标签与来源。

使用模型
GPT-Image-2
创作者
@Nine_Shell
发布时间
Thu Aug 06 12:59:35 +0000 2026
图片数量
4 张

完整提示词

源提示词按原始换行与标点完整保留
Made with Grok Imagine Video 1.5 (reference-to-video, 3 image slots).

All 4 source stills were generated with GPT-Image-2, simulating Fujifilm Pro 400H film. Images 3 & 4 (same subject, different framing — wide and close-up) were registered under a single reference slot for identity consistency.

Didn't fully think this through in advance — the 1→2 cut came out as a soft dissolve instead of a true hard cut, and there's a faint music bed under the ambience that I hadn't accounted for. Both are probably fixable with tighter prompt language. Might've felt even more like an actual film still if I'd caught those first.

Prompt used:

Cinematic 15-second film sequence, Fujifilm Pro 400H grain throughout, quiet wistful mood, soft overcast morning light. Open on @Image1, an empty Japanese shopping street at dawn, camera barely pushes in as power lines sway in the wind. Hard cut to @Image2, sunlight through a rain-streaked station window on an empty bench, camera drifts slowly across the glass. Hard cut to @Image3, a woman standing alone on a train platform in a wide shot, hair lifting in the breeze as she watches the tracks; the camera slowly and continuously pushes in until it settles on a close profile of her face, she blinks slowly, same face and outfit held consistent throughout. AUDIO: distant crow and wind over the street, rain tapping glass and a faint station announcement, then wind and a distant train bell fading as the camera closes in on her face.

提示词译文

译文与源提示词分开显示,不替代原文
使用 Grok Imagine Video 1.5 制作(参考生成视频,3 个图像槽位)。

全部 4 张源静帧均使用 GPT-Image-2 生成,模拟 Fujifilm Pro 400H 胶片。图像 3 与图像 4(同一主体,不同取景——远景与特写)注册在同一个参考槽位下,以保证身份一致性。

事先没有完全想清楚——1→2 的剪辑结果是柔和的叠化(dissolve),而不是真正的硬切(hard cut),而且环境音下还混着一层我没预料到的微弱音乐底床(music bed)。这两个问题大概都可以通过更精确的提示词语言来修正。如果当时抓住了这些细节,成片可能更像真正的电影静帧。

使用的提示词:

电影感 15 秒胶片序列,全片贯穿 Fujifilm Pro 400H 颗粒感,安静而惆怅的情绪,柔和阴天的晨光。开场为 @Image1,空荡的日本商店街在黎明时分,镜头轻微推入,电线在风中摇曳。硬切到 @Image2,阳光透过雨水划过的车站窗户照在空长椅上,镜头缓缓掠过玻璃。硬切到 @Image3,一个女人独自站在火车站台的远景中,发丝在微风中扬起,她注视着铁轨;镜头缓慢且持续地推近,最终定格在她面部的近景侧脸,她缓慢眨眼,全程保持同一张脸与同一套服装。音频:街巷远处乌鸦叫声与风声,雨水敲打玻璃以及隐约的车站广播,随后是风声与远处的列车铃声逐渐淡出,此时镜头推近至她的面部。
低饱和平视广角远景柔光怀旧街道城市或街道胶片或复古摄影角色一致性图生图参考图引导