I work for an architecture and urban planning firm.
And the image generation AI that many of us are looking for, first and foremost, is to keep the layout to CAD or BIM generated images and make it as realistic as live action with Gemini 2.0 flash (image generation) experiments and images/text.
However, no matter how we devise the prompts, it does not generate satisfactory images as shown in the bottom row, so please improve it!
And then next, after being able to generate it as realistic as live-action
For example, the image of a blue car driving on the road was good until the whole image was bright and soft, but when adding people, the number of people and where to draw them cannot be generated very well.
