AI integration ?

[Sonny Sonny]

Hi, does Serif have any plan for AI integration (E.g: stable diffusion, or Affinity built-in AI) in the near future?

[Westerwälder]

Das Thema hatten wir erst vor kurzer Zeit, aber wenn Affinity mit KI beginnen sollte würde ich nicht mehr mir Affinity arbeiten, denn die KI zerstört jede Krativität beim Fotografieren, weil die nichts neues zeigen kann, weil sie mit betehenden Daten gefüttert ist! Bildbearbeitung mit Automatiken produziert nur Mittelwerte!

We only had this topic a short time ago, but if Affinity were to start with AI, I wouldn't work with Affinity anymore, because the AI destroys all creativity when taking photos because it can't show anything new because it's fed with existing data! Image processing with automatic functions only produces average values!

[nomi02118]

I would ask what you mean by AI. I’m sue the topic AI photo app and plugin to denoise and sharpen and it doesn’t destroy my creativity. AI art creation is another story to me, it might be fun to play with but in my testing of it it does nothing to help me create art as Westerwalder might agree with. So for me as aid to make some tools more powerful I am for it but not for much else. Some of the stuff Adobe is doing like using it to make graphic fonts is an interesting idea but I also have the feeling they would all look very similar very quickly. 

[lphilpot]

Sharpening, healing, denoising, upscaling, etc. -- Sure. Creating? No way. As long as AI is a tool to carry out my wishes, under my control, that's fine. When it starts "creating" by itself, no interest at all.

[Komatös]

It remains to be clarified who owns the copyrights to the algorithm-generated images. Is it the author of the AI? Is it the user?

[walt.farrell]

1 minute ago, Komatös said:

It remains to be clarified who owns the copyrights to the algorithm-generated images. Is it the author of the AI? Is it the user?

In the US, the current decision is that no one owns them.

[AffinityMakesMeWonder]

Pretty impressed what online AI tools do these days…

Uploading three pictures (small part of a bigger pictures) that showing before and after AI sharpening/colorizing…

[Komatös]

11 minutes ago, AffinityMakesMeSmile said:

Pretty impressed what online AI tools do these days…

Uploading three pictures (small part of a bigger pictures) that showing before and after AI sharpening/colorizing…

AI-supported image optimization tools are, without a doubt, already a fine thing.

 

@walt.farrell

In Europe, there is a tendency for the authorship to lie with the AI developer.

Well, according to our former chancellor (Angela Merkel), the Internet is new territory. 😂

[nomi02118]

Tools like sharpening etc is something we could want easily to be included in future Affinity updates. I do a lotto old photo restoration and while these tools can have variable quality they can do a lot to making some photos usable or retired in a way manually is impossible so I am all for it. It's the AI creating the whole photo or illustration I think is the wrong use for the technology. 

[herhey]

I think AI is very useful for selections (subjects, sky, objects, etc)

[nomi02118]

Good point too, I should have mentioned that. That would be a nice addition to A-photo. 

[Sonny Sonny]

AI tools aren't bad on their own, but it's the people who use them that can make them seem that way.

[Affinityconfusesme]

Ai tools are very expensive to develop. They need to be written in python because python is made for AI. Adobe photoshop has these tools because they charge a subscription for photoshop which means they have the money to develop them.

[herhey]

There are a lot of good open source projects.

[ATP]

51 minutes ago, tzvi20 said:

Ai tools are very expensive to develop. They need to be written in python because python is made for AI. Adobe photoshop has these tools because they charge a subscription for photoshop which means they have the money to develop them.

As herhey mentioned there's a ton of open source projects which can do people and object recognition to a high standard of quality.

[jussi]

5 hours ago, tzvi20 said:

Ai tools are very expensive to develop. They need to be written in python because python is made for AI. Adobe photoshop has these tools because they charge a subscription for photoshop which means they have the money to develop them.

just to correct things a little, Python is just a wrapper using tons of various libraries. usually written in C++. however, these AI tricks require still lots of computational power and each of them require online cloud processing or sending data somewhere and receive the result. which might be a no-go for some users .. 

anyway, I would personaly also like some Affinity version of Neural filters currently available in Photoshop .. 

[ATP]

Some of the PS neural filters might be useful but things like Smart Portrait would be wasteful usages of Serif's resources.

I'd much rather some smarter tools to automatically select objects, people, skies or improvements to in-painting/content aware fill etc.

Cloud stuff is a no thanks for me, it's also expensive to run and locks us even more to Serif's servers than the internet activation introduced in V2 already does.

[Sonny Sonny]

Another simple way, I hope someone could develops an addon or a plugin that helps embed stable diffusion into Affinity UI, like some devs did for Krita, Photoshop, Gimp

[Intuos5]

On 5/2/2023 at 7:49 PM, herhey said:

I think AI is very useful for selections (subjects, sky, objects, etc)

Agreed, AI to simplify workflows to get something tedious done with 80% of the result and 10% of the work, otherwise, I see no use for AI tools.

With Photoshop's AI tools, you still have to correct the sky selections, etc, which fail to get non-blue skies, night skies, reflections, etc.

[Sonny Sonny]

While waiting for extension for AP, I'm using A1111 plugin for Krita and it works great! 

[Conrad Hoffman]

The only reason I keep Skylum LuminarAI around is because it can do some nice selective things with skies and color. Not sure how much actual "AI" is involved, but those sorts of tools are useful. I also use Topaz Photo AI, but it's very image dependent. It can wreak havoc with some types of images, though it can be good for restorations. In general, I don't want my editor to be creating or inventing anything, just remapping what's there. The metric I use is if the image were used in court, would the processing disqualify it from being a reasonable facsimile of the scene? IMO, anything processed through Topaz would be highly questionable. Manual cleanup of noise and tonal values much less so. If Affinity is going to add AI tools, they have to be clearly identified and not replace any more standard tools.

[Aisthetikos]

I'd like AI to replace the worst assembly line tasks that just steal time from creatives, like sharpening and selecting hair or correcting the color balance of entire images or selections.  Professional photographers will demand this so that they or their assistants don't have to bill clients (or themselves) for image processing like they do today.

I'm way behind in my own photo processing because I simply don't have the time or don't want to spend hours on image processing. I have taken many photos in my work and life, and many of them have potential if they go through image processing. But I never get around to it.

[Abelnt]

please ai generative fill soon or adobe will take my money again

i need right now all the news ai tools, sorry...

[Sonny Sonny]

15 hours ago, Abelnt said:

please ai generative fill soon or adobe will take my money again

i need right now all the news ai tools, sorry...

Or just API support for Automatic1111 integration! It would be awesome!