Text to Image AI Generators

[M. Robert Gibson]

Having heard about these text to image AI generators, and since there is an open source one names Stable Diffusion, I thought I'd give it a go at installing and playing with it.
Well, in the course of my research I found that there are already online, and free, text to image AI generators.

I'll give them a test run then.

Some of them require your email address, even though they claim to be free, but you are in effect paying for them with your data so you can be spammed and tracked and sold. So I didn't bother with those

These are the ones that are totally free (That I've tried so far. There will be others)
Craiyon - produces 9 images
DeepAI - allows a choice of style
AI Image Generator - allows uploading an image to guide the AI and even lets you draw something for guiding the AI
Dezgo - allows a choice of AI model to produce different results

For my tests I used two prompts

• A barbarian holding a sword
• A giant spaceship hovering over a city

The following posts will show the results

Edit
I've just found out that DeepAI only allows a certain number of free images, then it starts hassling you to sign in Shame because some of the styles were quite amusing

 

[M. Robert Gibson]

First up we have Craiyon

Craiyon - Your FREE AI image generator tool: Create AI art!

Craiyon is an AI model that can draw images from any text prompt!

www.craiyon.com

 

[M. Robert Gibson]

Next we have DeepAI

This one uses the Fantasy World Generator

Fantasy World Generator

Generate an image in fantasy style.

deepai.org

This one uses the Cyberpunk Generator

Cyberpunk Generator

Generate an image in cyberpunk style.

deepai.org

And these two use the generic text2img generator

AI Image Generator

This is an AI Image Generator. It creates an image from scratch from a text description.

deepai.org

 

[M. Robert Gibson]

Next is the AI Image Generator, without any uploaded image or drawing prompt

AI Image Generator - Free online AI art generator

A free AI image generator that lets you create AI art by providing text and image prompts to AI models like Stable Diffusion

www.aiimagegenerator.org

 

[M. Robert Gibson]

And finally, for now, we have Dezgo

Dezgo.com

Generate high-quality images from any text prompt. Let the AI draw!

dezgo.com

using the default DreamShaper6 model

 

[M. Robert Gibson]

And, just because I had to, I put my username, M. Robert Gibson, into Dezgo with the DreamShaper6 model

Uncanny! It's like looking in a mirror

But here I am according to AI Image Generator

 

[M. Robert Gibson]

So what's the practical application of all this?

Well, the thing that piqued my interest was wondering if the output could produce images good enough for the cover of books

From Dezgo using DreamShaper6 here's "a group of adventurers battling a dangerous dragon" Not bad if you don't look too closely

But if you do look closely, it looks like AI can't manage swords yet, and as seen from the other examples, hands are a big problem

 

[KiraAnn]

And these two use the generic text2img generator

AI Image Generator

This is an AI Image Generator. It creates an image from scratch from a text description.

deepai.org

View attachment 105735

And it looks like one of M Robert Gibson's examples needs more training in number of legs.

 

[OuttaInc]

So what's the practical application of all this?

Well, the thing that piqued my interest was wondering if the output could produce images good enough for the cover of books

From Dezgo using DreamShaper6 here's "a group of adventurers battling a dangerous dragon" Not bad if you don't look too closely
View attachment 105747

But if you do look closely, it looks like AI can't manage swords yet, and as seen from the other examples, hands are a big problem

If I see an AI-generated book cover, I'm going to assume the text is AI-generated as well.

Hard pass, for me.

 

[Omits]

I have tried Craiyon (just) but wonder if there is a rule for entries like ';' after input requirements. Or is it just narrative text like 'Alien spaceship no weapons dimensions 25L x 6w x 5h'?

 

[M. Robert Gibson]

From what I've read, Stable Diffusion converts the text into tokens, so I don't know if punctuation matters (other than spaces)

How does Stable Diffusion work?

Stable Diffusion is a latent diffusion model that generates AI images from text. Instead of operating in the high-dimensional image space, it first compresses the image into the latent space.

stable-diffusion-art.com

Tokenization is the computer’s way of understanding words. We humans can read words, but computers can only read numbers. That’s why words in a text prompt are first converted to numbers.

A tokenizer can only tokenize words it has seen during training. For example, there are “dream” and “beach” in the CLIP model but not “dreambeach”. Tokenizer would break up the word “dreambeach” into two tokens “dream” and “beach”. So one word does not always mean one token!

Another fine print is the space character is also part of a token. In the above case, the phrase “dream beach” produces two tokens “dream” and “[space]beach”. These tokens are not the same as that produced by “dreambeach” which is “dream” and “beach” (without space before beach).

Stable Diffusion model is limited to using 75 tokens in a prompt. (Now you know it is not the same as 75 words!)

I've just done a quick search and found this guide about punctuation

What Punctuation Does Stable Diffusion Use and Know? - Prompt Phantom

What punctuation marks does Stable Diffusion use for formatting? In this guide, we'll go over each mark that makes a difference in your AI art prompts.

promptphantom.com

 

[tinkerdan]

have you tried microsoft bing image creator from its Designer

from this prompt
A female space cadet, in a blue uniform with red trim, running forward holding a small ray pistol, with the interior of a space station in the background

these four images.
add
with blond hair and a cyan background with silver trim and red, green, blue, and yellow leds

change
with blond hair braids and a realistic style

 

[Christine Wheelwright]

Some of the space cadets have six fingers.

 

[tinkerdan]

Then just start making variations

@Christine Wheelwright , I count three with an extra finger-usually on the trigger.

 

[tinkerdan]

This one was--all the parts--generated in MS chat and assembled using Corel Photo-Paint.

 

[Lacedaemonian]

Some of the space cadets have six fingers.

AI cannot do hands.

AI art mostly seems slightly off to me but it is incredible in terms of creating concept art or piece that is later refined by the human hand.

 

[HareBrain]

AI cannot do hands.

I've seen evidence that it's getting better at hands. However, I've still not seen evidence that it can do people convincingly interacting (no, not like *that*, though that probably is what it will end up being used for).

 

[tinkerdan]

@HareBrain They can do some action in figures, however introducing more than one character is hard enough to accomplish let alone putting the action in them--not even touching upon their actions being interactive.
Part of the problem is getting the wording correct--although the biggest problem seems to be getting the AI to properly interpret larger stacks of information.

If you use more than 9 words to describe something it can begin to be confusing for the AI. You can add punctuation, however some punctuation can also mess the AI up.

Come to think of it I should ask if anyone has found good guidelines for text to graphic for AI and how to punctuate. I've noticed that these free image generators have little if any documentation.

 

[tinkerdan]

Here is another attempt in this one I tried to recreate the image below that I did in watercolors over 40 years ago.
I first imported the watercolor and had the AI describe it and then tried to use the description to generate the image.
I finally had to settle for the strange asteroid and the ship--which it seemed to do well enough; then I had it generate the mans image separately until I got what I wanted or close and then I fused those together with a background of stars.

 

[M. Robert Gibson]

Come to think of it I should ask if anyone has found good guidelines for text to graphic for AI and how to punctuate

Stable Diffusion prompt: a definitive guide - Stable Diffusion Art

Developing a process to build good prompts is the first step every Stable Diffusion user tackles. This article summarizes the process and techniques developed

stable-diffusion-art.com

There is a lot of useful information elsewhere on that site