Censorship removed this post and title

Prompting
,
38

Okay, things are clearing out.

Imagine this image as our start point;

image
image1792×1024 144 KB

If you want this image “but with a twist”, simply refer to it’s gen_id and change the seed.

But if you want the style / atmosphere, etc… and with a bigger change, add that change in your prompt and refer to the gen_id as the base image.

E.g. “show two woman with flip flops”;

image
image1792×1024 181 KB

As you can see the woman in the first picture (red dress) is also (kind of) available in the second image. And the second lady is swapped with the first man (same glasses / white cloths).

But the look and feel, park, stones, background, etc… are the same (atmosphere / style) because of the referred image.

You can now take the second picture as a starting point and use that unique gen_id for further iterations.

E.g. “swap the stones on the ground with grass”;

image
image1792×1024 205 KB

This is, somehow, the same image but now situated on a grassfield.

{
  "size": "1792x1024", // dimension of the image
  "prompt": "(text)", // text that describes the scene
  "n": 1, // number of wanted images
  "referenced_image_ids": ["Mrwfxmlt4MheddKh"], // the #ID of the original image
  "seeds": [3091385326] // the seed of this particular image
}

Choice matrix

  1. Want a complete different image?
    Write a complete new prompt.

  2. Want the same image style / concept, but with a twist?
    Refer to the gen_id and force another seed.

  3. Want the same image, but with some other details?
    Refer to the gen_id and write the changes in the original prompt.

Tip

When you refer to a gen_id, the same approach of that particular image will be used and also the same parameters (like prompt, dimensions and also seed).

So when you refer to a gen_id an say nothing about another seed, Dall-E will use the exact seed of the referred image (you don’t have to add it yourself).

3 Likes