Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Humans have nothing to fear from Adobe
- Thread starter dorsal
- Start date
- Sort by reaction score
Give Firefly a try, much better results. Not sure why they don't integrate Firefly into Photoshop rather then a separate AI service.
Go open source & free instead of commercial. Adobe's GenAI imagery can't actually compete for various reasons (amongst others because of legal copyright reasons).
These were rendered on my own local machine (not internet required) using different current AI models. For ~25% of these I added an "in the style of...", but otherwise the same prompt.
Took about 90 minutes to generate. If you think about it this it is maddeningly fast...
Gives a better feel for where GenAI imagery stands right now. And this is only the beginning. It is sometimes daunting to think about if you're an artistic person like myself. It will only improve and improve more. While it will not take away jobs from artists who stand out or when "hand drawn/painted by a real human" becomes a "seal of quality" for higher end products/art, average art jobs and lower level ones are already being phased out and replaced by GenAI... Video, voices, singing, music, 3d modeling, and so forth: same.
At this point hands still present problems to GenAI - but that is also improving rapidly. And yes: counting to three seems to be an issue Often the AI settles on 4 rather than 3 characters...
Full resolution webp one here: https://gofile.io/d/SMpzN2
These were rendered on my own local machine (not internet required) using different current AI models. For ~25% of these I added an "in the style of...", but otherwise the same prompt.
Took about 90 minutes to generate. If you think about it this it is maddeningly fast...Gives a better feel for where GenAI imagery stands right now. And this is only the beginning. It is sometimes daunting to think about if you're an artistic person like myself. It will only improve and improve more. While it will not take away jobs from artists who stand out or when "hand drawn/painted by a real human" becomes a "seal of quality" for higher end products/art, average art jobs and lower level ones are already being phased out and replaced by GenAI... Video, voices, singing, music, 3d modeling, and so forth: same.
At this point hands still present problems to GenAI - but that is also improving rapidly. And yes: counting to three seems to be an issue
Often the AI settles on 4 rather than 3 characters...Full resolution webp one here: https://gofile.io/d/SMpzN2
Go open source & free instead of commercial. Adobe's GenAI imagery can't actually compete for various reasons (amongst others because of legal copyright reasons).
These were rendered on my own local machine (not internet required) using different current AI models. For ~25% of these I added an "in the style of...", but otherwise the same prompt.
Gives a better feel for where GenAI imagery stands right now. And this is only the beginning. It is sometimes daunting to think about if you're an artistic person like myself. It will only improve and improve more. While it will not take away jobs from artists who stand out or when "hand drawn/painted by a real human" becomes a "seal of quality" for higher end products/art, average art jobs and lower level ones are already being phased out and replaced by GenAI... Video, voices, singing, music, 3d modeling, and so forth: same.
At this point hands still present problems to GenAI - but that is also improving rapidly. And yes: counting to three seems to be an issue
Full resolution webp one here: https://gofile.io/d/SMpzN2
Impressive. What AI tool did you use?
Stable diffusion locally installed with a browser-based GUI. Then download AI models, Loras ("extensions" that are trained for special purposes, such as a particular actor or a style like pixel art), and embeddings (simple keyword based extensions, I believe) from https://civitai.com/
Setting up a local installation may be super simple to quite complex, depending on your skill level and the approach used.
Here is a good article:
decrypt.co
For example, a very simple artist-friendly approach is to install Krita on your Mac
krita.org
(which is the brilliant open source and free digital painting app)
and then simply install a free plugin:
github.com
And Presto: you have Photoshop's "Generative Fill/Expand" for free and without any limitations in terms of content. And far more controllable! And better quality. And super easy to use within a professional digital painting app. All for free!
The genAI community hub right now seems to be https://civitai.com/
That's where I downloaded many different models.
Setting up a local installation may be super simple to quite complex, depending on your skill level and the approach used.
Here is a good article:
The Best Ways to Run Stable Diffusion and SDXL on an Apple Silicon Mac - Decrypt
In the same way that there are workarounds to achieve high-powered gaming on a Mac, there are ways to run Stable Diffusion.
decrypt.co
For example, a very simple artist-friendly approach is to install Krita on your Mac
Digital Painting. Creative Freedom.
Krita is a professional FREE and open source painting program. It is made by artists that want to see affordable art tools for everyone.
and then simply install a free plugin:
GitHub - Acly/krita-ai-diffusion: Streamlined interface for generating images with AI in Krita. Inpaint and outpaint with optional text prompt, no tweaking required.
Streamlined interface for generating images with AI in Krita. Inpaint and outpaint with optional text prompt, no tweaking required. - Acly/krita-ai-diffusion
github.com
And Presto: you have Photoshop's "Generative Fill/Expand" for free and without any limitations in terms of content. And far more controllable! And better quality. And super easy to use within a professional digital painting app. All for free!
The genAI community hub right now seems to be https://civitai.com/
That's where I downloaded many different models.
I will mention though that these stable diffusion models are generally trained on contents scraped off the web without any thought given to copyright or the rights of the original artists and photographers.
In short: play as much as you want, but the results are neither morally nor ethically acceptable for any paid work in my opinion. The legality is still very much in the dark here. No-one knows quite how to handle this part.
One of the reasons why Photoshop's implementation is so limited. And Firefly refuses to render certain (common and seemingly safe to use) keywords.
In short: play as much as you want, but the results are neither morally nor ethically acceptable for any paid work in my opinion. The legality is still very much in the dark here. No-one knows quite how to handle this part.
One of the reasons why Photoshop's implementation is so limited. And Firefly refuses to render certain (common and seemingly safe to use) keywords.
Interesting stuff!Stable diffusion locally installed with a browser-based GUI. Then download AI models, Loras ("extensions" that are trained for special purposes, such as a particular actor or a style like pixel art), and embeddings (simple keyword based extensions, I believe) from https://civitai.com/
Setting up a local installation may be super simple to quite complex, depending on your skill level and the approach used.
Here is a good article:
The Best Ways to Run Stable Diffusion and SDXL on an Apple Silicon Mac - Decrypt
In the same way that there are workarounds to achieve high-powered gaming on a Mac, there are ways to run Stable Diffusion.
For example, a very simple artist-friendly approach is to install Krita on your Mac
(which is the brilliant open source and free digital painting app)Digital Painting. Creative Freedom.
Krita is a professional FREE and open source painting program. It is made by artists that want to see affordable art tools for everyone.
and then simply install a free plugin:
GitHub - Acly/krita-ai-diffusion: Streamlined interface for generating images with AI in Krita. Inpaint and outpaint with optional text prompt, no tweaking required.
Streamlined interface for generating images with AI in Krita. Inpaint and outpaint with optional text prompt, no tweaking required. - Acly/krita-ai-diffusion
And Presto: you have Photoshop's "Generative Fill/Expand" for free and without any limitations in terms of content. And far more controllable! And better quality. And super easy to use within a professional digital painting app. All for free!
The genAI community hub right now seems to be https://civitai.com/
That's where I downloaded many different models.
I downloaded Diffusion Bee in the first link, and then tried downloading a model from the Civitai website, gave me a big error message. The Diffusion Bee says the extension is supported. Is Diffusion Bee maybe a more basic app and something more extensive is needed?
Attachments
It looks like you tried installing a diffusion checkpoint that is unsupported by Bee. Avoid an XL diffusion model, and try a checkpoint such as Realistic Vision or DreamShaper. PS I have no experience with Bee myself.
Also try the Krita plugin: I haven't experienced issues with that one. Simple to install and the plugin automates the download of the most important AI models, so you can get started after a few minutes.
Also try the Krita plugin: I haven't experienced issues with that one. Simple to install and the plugin automates the download of the most important AI models, so you can get started after a few minutes.
Last edited:
So I installed Krita and the plugin for AI. I have a connection error that I can't figure out. In the video I watched about installing this there was zero issues with this. I went into config but have no clue what I am looking for. Any thoughts?
**edit, never mind I got it figured out!
Attachments
Last edited: