AI Drawing (Midjourney etc.)

WHAT DO YOU RECOMMEND

It would be a good idea to introduce AI drawing like midjouney and implement them in Anytype.

HOW COULD IT BE DONE

ᅟUsers can directly generate desired pictures according to their own requirements and place them on the page or mark, which can save users the time and cost of going online to find a desired picture.

REAL WORLD USE CASES

ᅟDiscord has already started, although it is not a note-taking application, but I think it is still worth learning.

RECOMMENDED ALTERNATIVES

ᅟAs mentioned, AI drawing can be combined with AI assistant.

1 Like

Just my opinion, but isn’t the copyright issues brought on by AI art-generating tools misaligned with the app’s goal/claim of being trustworthy?

That said, people are free to upload whatever they want and create plugins. :tipping_hand_woman:t2:

Cost-wise, though, I don’t think it’s going to be any better. Decent AI art (like those from Midjourney) takes time and resources to refine, and usually require a subscription. (Let me know if I’m wrong, as I don’t know how they do it in Discord.)

4 Likes

Midjourney doesn’t seem to have a public API at the moment, but DALL-E does have one.
Also, there is stable diffusion which is free locally and mostly free on the web trough third-party apps or websites.

1 Like

DALL E uses tokens/credits. I did a bit of reading and it sounds good from a consumer’s perspective, though a lot of reviews say it’s still not on the level of Midjourney when it comes to quality.

I am wary of Stable Diffusion. Personally, the first thing I look at is how these ai art generators handle art theft. If I don’t find anything addressing artists wanting to opt out of their image bank/database, then I’m just so wary of supporting the company. In their website, they said there was no opt out for their LAION 5B dataset, which is what they used to train their ai. (Basically copyrighted images)

I’m all for giving artists the option to opt out of the datasets (or better still, have it opt-in only) but don’t tell people that that’ll protect their work from being replicated by AI. There are millions of images that are out of copyright and all of us learned to draw and paint from other people. The correct recipe of old masters and photographs can give you pretty much any image when it comes to how the diffusion models work.

I’d also be extremely cautious of adding native AI support to AnyType. As an external API that is able to act only within a siloed section of your app, maybe. Don’t forget that things like Midjourney and ChatGPT ingest everything you give them to improve their training and this means that you’re content is compromised by default.

2 Likes

Hi. So I’m not sure if this is in reaction to my comment or the general topic, but I just want to clarify that I don’t mean to say that supporting artists will prevent AI from learning how to replicate work. The tech is already out there and I know it’s improving at an alarming rate. My point is just to consider the ethical side first before joining the AI bandwagon.

3 Likes

:pencil2:ᅟUsers can directly generate desired pictures according to their own requirements and place them on the page or mark, which can save users the time and cost of going online to find a desired picture.

Lets talk about that part “time and cost”. What is the time and cost of generating an image at a decent resolution with Midjourny. How many times will have you have to try again and vary the prompt to get what you actually want ?

Compared to having a UI hooked up to some image searching service like unsplash or google image search.

I get that AI is currently hot and exciting but that does not make it a good solution to every problem.

3 Likes

Once again plugging Tana. They have a integration with DALL-E in addition to other OpenAI integrations.

This is such an important facet of AI and people rarely talk about how this impacts privacy and security. In time it is possible that users will have no idea how their information is being used by the AI because at some point the AI begins writing its own code.

1 Like

For Midjourney and similar at least, aren’t you just giving them prompts? There doesn’t seem to be any other data involved.

@Filip I think you are right. My comment had more to do with ChatGPT. I believe you are correct that Midjourney is prompt base, which limits what someone is likely to share.

1 Like

It’s possible to upload images on Midjourney.

This article talks about the complications of AI art.

1 Like

I wrote this in another topic but wanted to leave it here also.

Current state of regen"AI" is that it was made using civilisational level of IP violation and poses unforseen amount of privacy and security risks (not to mention negative economical turmoil it was designed for). It is in direct conflict with Anytype phylosophy, statements and goals as I understend them.

Adding this to the fact that I am a guy who is professionally connected to 2&3D creation would mean a instant deleting any of my data related to Anytype.

1 Like

As an artist I’d like to voice my support for Anytype to NOT implement this…there’s so much art theft it’s ridiculous.

1 Like