Tiny Mouse Adventurer: Can you spot the problems?

[Rodney]

Many of you have seen my Adventure Cat drawings but this is a little different...

There are several problems with this image.

Perhaps you can spot them?

 

[robcat2075]

His tail is floating the air?

Cute image!

Is that a 2D painting or 3D Model?

[Rodney]

Good eye.  I didn't see that!  (extra points to you)

Here's a variation on the theme that has his tail more on the ground:

 

Disclaimer:  I'm not giving away much about the image(s) yet!

[Roger]

Is this an AI image?  Because his left paw looks like it has too many digits.  Likewise, the right paw looks too much like a human hand. 

Otherwise, it's a cute concept. 

[robcat2075]

Between the two, it's not clear to me if he's standing ON the log or behind it. His toes are in an awkward tangent with the edge of the log.

I don't like the cauliflower ear in the second one.

[Rodney]

@Roger gets the prize.

Both of these images were in fact created using 'AI' text prompts.

 

The 'problems' are manifold (err... many-fold).

 

Problem 1: Authoring

This is a bit like googling for an image and then claiming the resulting images as something we've created ourselves... potentially quite problematic.

(I do like the little mouse though!  And this was the first image created... all subsequent images where of lesser interest/quality)

 

Problem 2: Hands and Hoverings

Historically AI generated images have been notorious for not being able to create 'hands' and...

This likely fits into @robcat2075's observation of things (tails) floating... AI images often have shapes that are suspended in place without anything to ground them or connect them to other shapes.

I see this as the difference between real 'intelligence' and the smoke and mirror of patten matching/prediction that easily reads as intelligence... in many case indistinguishable even.  There is no doubt however that speed is a huge advantage for these generative processes.

 

I'm sure there are many more problems but I've been playing with AI of late in a variety of ways... creating images, music, stories, code... and find it all quite useful.  In many cases VERY useful.

And yes... problematic.   

 

[robcat2075]

I like the look of the character and one could absolutely make that in A:M but if I Google "adventure mouse" I will find existing images that have the key elements present.

If I made that character I'd worry that I was borrowing too much.

[Rodney]

Mouse characters are pretty well covered in cartoon land so anything created (or imagined) is likely going to be seen as derivative.

That'd be an interesting challenge... create a mouse character that is 'truly unique'.

 

It'd be interesting to do some comparisons with standard google searches.

My quick look turned up a lot of mice... some more adventurous than others.

None particularly looked like the one created in this instance.

 

An unanswered question I have would be; what becomes of the generated images made with AI?  Do they get stored and referred to for subsequent generating processes?

I must assume so.

 

I've created many mouse characters over the years... some certainly more interesting than others.

 

The process used to generate images via AI prompting is an interesting one... and one we tend to do in similar fashion ourselves when we create such things.

The big difference being that (supposedly) we are intelligent. 

 

Immediately after seeing the generated image I asked myself the question; if I had drawn/painted/sculpted this what might I do differently.

My very first thought was 'add a backpack'.  (i.e. replace the current backpack with something that reads more as a 'backpack')

Second thought; 'add subtle flowers in the background'.

There is certainly more that could be changed but I'd probably leave that for after developing more of the story.

The story being the driving force for any changes worth pursuing.

 

 

 

[Rodney]

Here's another mouse for you...

The prompt:  photorealistic, tiny mouse, camera down angle, solid white background

 

My bad.  I didn't properly convey that bit about the camera. 

[Rodney]

Take 2:

Prompt:  photorealistic, tiny mouse adventurer, camera view is down angle on mouse, solid white background

 

Note the extra toe on the left foot.

[Rodney]

There are a lot of ways (and services) available for generating AI images.

There is also just google as well (no kidding).

 

If anyone hasn't done the AI image generation thing and would like to many services are using "pollination.ai" under the hood.

I believe in turn they use Stable Diffusion.

 

 

It's certainly not required but as a programming exercise I created a simple C++ console program that asks "What kind of image do you want to create?" and then launches the URL lookup that generates the image.  It then returns in loop to ask if we want to create another image.   A plus here being that with the console program we can arrow up and down through previous promptings and adjust them as necessary.

A more direct generation can be made by going to their web generator:   Pollinations

 

And yes, this is all research for eventually creating a tiny white mouse adventurer in Animation:Master.