How To Adapt Existing Images Using DALL-E

Jon Manderville
November 16, 2023

✅ You see or have just created the perfect image for your site/newsletter or Blog post.

✅ You used a tool like Midjourney maybe (one of the new wave of great timesaving AI helpers we can leverage to make light of creative work). 

🤔What if you want to make a variation, a wider version, a smaller version, the colours aren't right or the central image them is 'ok' but you would like to tweak it?

Midjourney has some tools to help you in the initial prompt.

Using "--seed [image ID]" is a great one and using the adapt/variations and upscale buttons when you get an image you like are all really good for the majority of the simple use cases. 

But what if you didn't create the original, or you want to combine 2 existing images to create new artefacts? Copyright laws need to be respected of course but with AI you now have a fantastic new fantastic option. And it lives adjacent to ChatGPT - its called DALL-E

You can access this same capability via Windows Copilot (it fronts up for DALL-E) but you get to have a much more specific interaction if you want to go in via Open AI's site. You can learn how to integrate ChatGPT with Power Automate to perform AI generated responses as part of our Power Automate PathFinder.



An Example Original

Try this alongside me if you like. See what happens for you and how you like the process

Imagine you like this image. 

dall·e 2023 11 13 09.36.46 create a cartoon style isometric image of a single person and an ai robot working together on a workflow grid. the person is using a laptop, and both

Maybe you created it but now you want to change the colours, the theme and perhaps the message you are giving.

Right now, for me this image says, "AI assistance for a task". I get the Microsoft colours to perhaps its about Microsoft Products. 

Lets try some changes 

Go to either DALL-E via the Open AI site or use Windows Copilot (I'll stick with Dalle for now for consistency).

Copy the image above (take a snapshot, save it, whatever you like. We aren't precious)

Paste it into the prompt and now ask: "I want the main background to be more muted and the character to be larger on the image"


I've now got 2 totally new images influenced by the first one that I can use. 

What if I want to change proportions?

No problem. If you want to fit any of the images to perhaps a square location (300x300), this sort of change used to be a bit painful unless you were pretty ok at image editing and tools like Adobe illustrator. How about we have a go now with DALL-E. 

Try choosing an image and typing "change to a square image 300 pixels by 300 pixels"


What I LOVE about this interface is I'm not just getting random errors spat back at me. I'm getting helpful suggestions that prompt me for more consideration. 

 Perhaps an aspect ratio change might be better?


ok, so what we are seeing here is that "editing" per-se is not possible but creating something new from a version we have saved is something in the DALL-E wheelhouse.

An observation I made at this point is that if I want to revert to the original, I would have to restart the conversation from that point. What seems to be happening is we are getting a new image influenced by the changes we have made to date. 

What If I want to change the main characters, or merge a new one in?

As a final test about what's possible for really common use cases, how about trying this:

"change the person character to a cat and the robot to a computer"  


I did try to introduce our logo but was told DALL-E "cannot assist with requests that contain logos or specific copywrited material".  I'm pretty sure I could introduce new elements quite easily that maybe were on brand with our logo so that's a change I can work on later. 

Aside from changing colours, and dimensions, why not try :

  • Merging 2 images to create a new one. 
  • Changing the 'feel' from a cartooon to photo realistic. 
  • Using tone and theme wording to change the overall impact of your image and the story it tells?

Seeing what I can do in 2 mins, we can already see its quite different to the original - maybe better, maybe worse. You decide.

Also its very much about the journey you want to take with your images. If its just applying a logo, perhaps use the tools build for layering but if its theme, tonal change and creativity you want. Give DALL-E a try.

We'd love to see what you can do too. Why not go ahead and experiment. Send us your results in the Forum within the Academy. We love seeing how people are using AI (don't forget to include your prompts so others can learn too)

Join 11,000+ in the Collab365 Academy

Master Microsoft 365, Power Apps, Power Automate, Power BI, SharePoint with Exclusive Access to 450+ Hours of Expert Training and a Wealth of Resources!