You see a picture on social media that makes you marvel. It could be the backdrop for a launch party or aPoster Design.
But trying to create a similar style is difficult.
At work, the boss or client sends a reference drawing and wants to design in the style of the reference drawing.
But one can do nothing with it.
In the past, we could only accomplish these needs through PS, adjusting the background, light and shadow, color tone and other parameters.
However learning PS is time and energy consuming, and even after learning most of the features, you may not always be able to realize the results.
While AI has brought us many changes already, there is still a huge chasm between a precise AI cue word, and a beautifully crafted picture:
Like knowing what the scene is, but not the style;
Knowing where the subject is in the picture, but not knowing what composition to use;
Knowing that the figures in the picture are decorated, but not depicting the characteristics of the image;
Constantly adjusting the cue words and still not achieving results;
…
Some might say, I'll send the image to AI for a one-clickInverse imageCue word.
This is indeed a good way to go about it, but analyzing is analyzing, and it seems that most of the time it doesn't work as well as one would like:
Is there any way to do this once and for all?
Here's a thought:
Before you achieve a goal, ask what AI needs.
I would like to upload an image and then you can analyze it based on the image information so that I can generate an image with a similarity of 95% in Instant Dreams 3.0, what information do you need?
As you can see, the necessary information to generate an image with a similarity of 95% is:
Main object / scene environment / color system / composition structure / artistic style / texture details / special effects elements
Based on this, we then backtrack the cue word:
Please explain in detail the content of the picture, including the main object / scene environment / color system / composition structure / artistic style / texture details / special effects elements, etc., carefully depicting all the design techniques of the picture, backward to me complete that the Dream 3.0 AI drawing Chinese cue words, so that I can directly generate 1:1 images in the Dream!
Take this graph above and test the results
Get a cue phrase:
Interior scene, a long table full of all kinds of fresh fruits, including watermelon, apples, pears, grapes, oranges, pineapples, etc., colorful and full. Soft natural light shines into the room, creating a warm atmosphere, with light-colored curtains and potted green plants in the background. The text "Chinese consume the world's watermelons 60% apples 49% pears 73%" is highlighted at the top of the screen, and the source of the data is labeled "Food and Agriculture Organization of the United Nations (2023)", and the overall style is realistic and delicate The data source is labeled "Food and Agriculture Organization of the United Nations (2023)", and the overall style is realistic and delicate, focusing on the texture of fruits, light and shadow levels, and conveying the data information with life-like scenes, with the logo of "AI drawing" added at the bottom
After removing the extra text, drop it into the i.e. dream to generate it:
The scenes are largely restored.
In this area of text try to still remove it all and just keep the background for a better result.
Test a couple other graphs:
Same, get a full cue word. (m² is not recognized, only the text is put in)
With the above cue words, you can basically complete a similarity of 90% or more images.
Well, that's all for this issue!