AI, AI, AI, wotz goin on 'ere then?

[toprob]

Here's another of my landscape photos with an AI-added carousel.

Have you ever wanted to revisit a childhood memory and see it with fresh eyes? With the new Bing image creator, you can do just that. You can use this tool to create realistic images from simple descriptions, such as "an old, rusty abandoned carousel in a field". This was one of my most vivid memories as a kid, when I stumbled upon this mysterious and magical sight while exploring the countryside. I felt a sense of wonder and curiosity, mixed with a hint of sadness and nostalgia. I wanted to know the story behind this forgotten relic, and imagine what it was like in its glory days.

Using the Bing image creator, I typed in my description and waited for a few seconds. The result was astonishing. It looked exactly like how I remembered it, but with more details and colors. I could see the intricate patterns on the horses, the faded paint on the poles, the cracked glass on the lamps. I could almost hear the music playing and the laughter of children. It was like traveling back in time, but also seeing it from a different perspective.

The Bing image creator is more than just a fun toy. It is a powerful tool that can help you explore your memories, emotions, and imagination. You can use it to create images of anything you want, from realistic scenes to fantastical landscapes. You can also share your images with others, and see what they create from their own descriptions. It is a great way to connect with yourself and others through visual storytelling.

This morning I started up my browser and it told me that there were some new tools available, so I took a look. One is the AI image creator, an AI image creation tool based on DALL-E. So of course I fired it up and asked for an image of a rusty old carousel.
By the way, the quote above didn't really exist until I asked the new Bing Compose tool to give me a starting point for this post. It just made up the bit about my recollection of the carousel...

Here's an image using the prompt 'a rusty, abandoned carousel sitting in a field'.


Here I added 'New Zealand landscape' which gets a bit closer to what I wanted...


And the same prompt but as a pencil sketch.


Ok, I prefer my own photo at the top, but the others are great for a few seconds' work. The AI creator tool gives square images, but I used the Photoshop AI to extend them left and right. Of course, anyone would...

I did ask the Compose tool for a post about exploring autistic obsessions using the AI image creator, but it took a dark turn. In reality, this is just a normal obsession.

Anyone else used the new Bing tools built into MS Edge?

[Charl]

Mmm the perspective in #1 is wrong.
You'd want to add something about distance to the prompt.

[cowpatz]

Incredible lighting and shadows.

[toprob]

Mmm the perspective in #1 is wrong.
You'd want to add something about distance to the prompt.

The top pic uses Photoshop Generative Fill, which began with my own photo, whereas the others are all AI-generated using Bing Image Creator.

With Generative Fill, I draw a mask -- which can be any shape, really, but in this case it's a rectangle -- where I want the prompted subject placed. The size and placement of the mask determine what the AI thinks can fit. If I move the generated subject, I can 'regenerate' to make it fit the new location. So to tweak the result I can move the result up or down and regenerate, which effectively moves it forward or back in the scene. If I move it really close, without changing the size of the mask, it'll know that it can only fit a part of a carousel, so it'll place a couple of horses or a bit of broken frame. Normally when I am tweaking the result I would change the size of the mask to match the new distance, but this is very relative.

In this case, I didn't want the carousel to be the subject, just one part of the landscape, so this under-emphasis worked best for me. Plus, every time you regenerate, you have no control over what you'll get, shape/colour or even subject-wise. So getting a pleasing appearance here was more important than anything else. You have to know when to stop..

With Generative Fill, the shape of the mask has more effect on the result than the prompt, really. So it is fun to experiment with wide/tall masks, blobs, and mask placement, but a big limitation is that if you get the ideal generated object, you can't move it or change it in any way because any regeneration will lose your perfect result. In that case you can either keep regenerating (each time gives you three options to choose from) or just revert to your best result, even though it might not be placed ideally. (All previously regenerated results are available unless you delete them to reduce the file size.)

'Generative fill' is quite a mouthful, sorry, I wish they'd called it something else.

[Splitpin]

Amazing stuff Rob ....but something a little spooky, but I keep looking at it .

[Splitpin]

Rob, thought i would give it a shot ....blue flying snail ....and...


what an age we live in , this will keep me at home all day

[Charl]

Marty...you said it: holy cow
And he said it, too:

You have to know when to stop...

Now my next question is this:
At what point do we get an in-sim airshow crowd for the Ardmore MSFS scenery???