Conversational Editing Workflow — Final Balanced Campaign

Conversational Editing Workflow — Final Balanced Campaign

原始文档

GPT Image 2 · en

Scene:
Square crop of the final version of the same transit campaign poster on the same review wall, straight-on and evenly lit. It should read as the polished result after several conversational edit turns, with a clean professional finish and no visible trace of the editing process.

Subject:
Preserve the original poster identity from the reference: central tram shelter, repaired empty left side, retained right pedestrian, abstract enlarged heading area, dusk environment, lower information block rhythm, route motifs, and square campaign framing. Make the final version more balanced and presentation-ready: refine spacing, clarify the hierarchy, tune contrast, and make the cool palette elegant rather than flat. The poster should look as if a designer accepted the conversational edits and produced a finished campaign proof.

Important details:
This image must demonstrate cumulative instruction following. The earlier requested changes should all remain visible: dusk atmosphere, no left pedestrian, larger abstract heading, cooler overall color, and preserved base composition. Rebalance only through subtle graphic design decisions: cleaner negative space, sharper icon-like abstract route marks, smoother illustration edges, and better tonal separation. Do not invent a new poster, change the topic, or add literal explanations.

Use case:
Conversational Editing Workflow series study 4/8, final cumulative edit proving that multiple conversational revisions can be resolved into one coherent image.

Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no left figure returning, no warm midday lighting, no software UI, no before-after split, no annotation labels, no brand identity, no chaotic poster elements.

使用方法

  1. 复制完整原始提示词,并保留其中的参数。
  2. 按需求替换主体、文字和构图要求。
  3. 在兼容的外部工具中使用,核对当前使用条件并检查结果。

开始前需了解

  • 先确认是否需要参考图。标注兼容模型,不代表该示例由这个模型生成。
  • 核对名称、数字与画面文字;不同模型版本的结果可能不同。