Skip to content

Keeping a consistent scene in ChatGPT & Dall-E when rendering images

When you are rendering images in ChatGPT (using Dall-E) you can ask for the seed of the image to work with.

Ask for the seed!

The seed allows you to you get roughly the same scene to use as a base.

You can then adjust the image with more prompts but you won’t lose where you are up to.

Here are images Is have created with this prompt:

PROMPT: Please generate an image of a mouse who is a knight riding his hamster. Make it cute and whimsical.

I received two images and chose this one:

I then asked for the seed of the image – which turned out to be 408728739.

Then I prompted for for the original image to be rendered in different styles.

PROMPT: using the seed in the second image regenerate the image to be a 3d render

Lets get working:

PROMPT: using that same seed – make the image a watercolour

PROMPT: using that same seed from the original image – make it an oil painting

PROMPT: using the same seed make the image a cartoon style

Notice how the seed number appeared on the bottom of this image – interesting!

So asking for the seed number of an image allows you to work with the same scene and only adjust certain aspects. Giving the seed as a limitation actually makes rendering images easier to work with!


Note – I use the paid version of ChatGPT for this rendering. It costs $20USD per month ($30AUD). As I use it for coding daily – it is an amazingly cost effective investment.


Leave a Reply