All language subtitles for 3. Comparison of AI Art Generators

As a generative artist, you should know the alternatives for Adobe Firefly.

There are many other tools in the industry, but in this video I want to compare the most popular ones.

Namely, we will be talking about Firefly, Midjourney, Stable diffusion and Dall-e.

There is no best pick.

Each tool has advantages in different areas.

The best way to figure out which one works best for you is to try them out and see which fits your workflow

the best, and I will try to explain their differences.

Pros and cons according to my research.

Let me introduce the tools first.

Starting with Dall-e.

This AI tool was launched by OpenAI in 2021 and it quickly gained attention due to its ability to generate

unique images from textual descriptions.

The software's name is a mix of the names of animated robot Pixar character Wall-E and the Spanish surrealist

artist Salvador Dali.

So Dudley is the first mover Mid-journey launched in July 2022.

This art generator was designed by a team of AI enthusiasts who aim to make AI art accessible for everyone.

Midjourney is currently only accessible through a discord bot on their official Discord server.

You can direct message to the vote or invite the bot to a third party server.

Stable diffusion is a deep learning text to image model released in August 2022.

Its unique feature is the capability to gradually evolve the artwork in real time, allowing the artist

to observe the creation process and make alterations at any stage.

It was developed by the Startup Stable in collaboration with a number of academic researchers and non-profit

organizations.

Last but not least, Adobe launched Firefly in March 2023.

Releasing the beta text image allows users to input a natural language prompts and receive an image

similar to other generative AI image models.

But there are more features in Firefly.

At launch, the Firefly beta is only available through the website, but Adobe stated it will soon be

integrated across its apps and it is already integrated with Photoshop.

All of these models are trained using a dataset.

Dall-e was developed with something called clip.

It is by OpenAI.

Clip is a separate model that was trained on 400 million pairs of images with text captions scraped

from the internet.

Midjourney uses many datasets like Lion, which was scraped from the web as well.

In Forbes interview, Midjourney founder David Holtz has admitted that Midjourney did notrillioneceive

consent for the hundreds of millions of images used to train the AI.

Stable diffusion is no different.

It again used pairs of images and captions taken from Laion five for training, which includes 5 billion

pairs of images and text.

The dataset was created by Laion, a German non-profit which receives funding from stability AI.

As many times stated in this course, Adobe Firefly trained on a dataset of Adobe stock and they are

saying that they did not train on any creative cloud subscribers personal content.

Next up, capabilities.

In terms of capabilities, there is no official way of comparing these tools.

Besides, they are better in different areas.

So I am going to share my own thoughts based on my own experience.

So I decided to say the key features of each tool.

Dulli is the first mover and still good for photorealistic images.

Also have API access which lets you create new images in any number of different styles as specified

by the user's prompts and include in your projects.

Mid-journey, in my opinion, is the best in terms of image quality.

As you might guess from its popularity, the best output comes from Mid-journey.

For example, Mid-journey is the only model that can handle hand generation, while others still struggles

with hand drawings.

You can define different artist styles, genres and mediums.

However, Mid-journey always creates images in in Mid-journey style and you can easily differentiate

the image generated with mid-journey.

Somehow it generates aesthetically pleasing results.

There is no API or standalone platform for mid-journey at the moment.

Only way to use it is through Discord app.

Stable diffusion is best for developers since it is open source and can be run locally on your own computer,

which means your system capabilities is important as well.

For example, ten gigabyte or more Ram is recommended.

On the positive side, stable diffusion basically gives you more options and control.

You can set the number of steps, the initial seed and the prompt strength.

Lastly, Adobe Firefly is really competitive when it comes to image quality, but there is more than

text to image like text effects and generative fill.

Generative fill is one of the most revolutionary thing about Firefly, since it opens a new door in

AI art.

It will be integrated in creative cloud programs like it is already been in Photoshop.

But I also asked the Firefly team if it's going to be a standalone platform and the answer is not clear

at the moment.

They said they are still testing and considering the roadmap.

All right.

That was the wrap up of capabilities.

And please consider that ecosystem is changing fast.

By the time you watch this video, some things may have already changed.

Lastly, let's see the pricing models of these tools.

Dall-e gives you 325 credits.

There is a rolling monthly 15 free credits allowance.

If you need more, it costs $15 per 115 credits.

Midjourney is a subscription service with prices starting at $8 per month annually and $10 per month

monthly.

Stable diffusion is free to use on the web app and you can run the tool without paying as well.

But if you want premium advantage, there are package starts from $9.

Adobe Firefly can be accessed for free using the beta access to use Photoshop beta.

However, you must be a creative cloud member with a subscription or trial that includes Photoshop.

All right, guys, that was the comparison of AI art generators.

Thank you for watching.