/biz/ - Business & Finance

[Anonymous No.53844248]

Anons, gib me a model on how to monetize Stable Diffusion.
I've spent about 3 months learning the ins-and-outs, and I want to make a profit so I can buy more precious metals, crypto, and hopefully launch a legitimate business.
I already have a few ideas, but I want to know what you anons have to say first.

[Anonymous No.53844279]

I'll pay you to stop spamming the general with unending waves of mid-tier monerochans with resolutions so low you can't even appreciate them.

[Anonymous No.53844295]

>>53844248
>generate a few different looking women
>generate different poses and contexts of these women, make it sexual
>create onlyfans accounts of these women
>use an teenage girl ai chatbot to generate text responses to simps
>????
>PROFIT!!!!!!

[Anonymous No.53844310]

>>53844248

>learning the ins-and-outs,

And by this you mean you've read a bunch of machine learning papers to understand the actual mechanism by which these algorithms work, or by ins-and-outs, you mean you've been generating thots, anime, and sci-fi images 24/7?

If it's the latter, kys immediately.

[Anonymous No.53844359]

>>53844279
Alas, I'm not the genius that got that racket set up.

>>53844295
I guess I could train a LORA of a specific look of a woman, and have that be an account.
Thanks, anon.

>>53844310
I did both. Started with anime girls, and ended with learning about diffusion algorithms.
I am not an engineer, and I'm not going to pretend I know everything about it, but I bothered enough to learn Python and the fundamentals of ML with it.

[Anonymous No.53846209]

bump

[Anonymous No.53846390]

>>53844359
You got any resources/articles for stable diffusion? I was recently googling how to generate a person, then keep that person and put that person into different clothing, accessories, and contexts. I found a great article but closed my browser and can't find it again. What is this LORA you speak of? I know DL/ML, but haven't really looked into stable diffusion. I think it would even be possible to create an IG profile and get lots of views and if you can generate an image of a girl wearing products, start to market your IG as an influencer to brands and shit, collect that advertiser money.

You could probably analyze top trends and what makes a highly viewed and liked profile/post through statistical analysis too.

[Anonymous No.53846912]

bomp

[Anonymous No.53847454]

bump

[Anonymous No.53847535]

>>53846912
>>53847454
Answer my post jerk.

[Anonymous No.53848391]

>>53847535
It's a different anon, anon.
I'm OP, sorry, I wasn't home, was buying some stuff at the store.

A LORA is a Low Rank Adaptation, basically a hypernetwork (a miniature "model" so to speak) you can use to generate a set of specific features - much easier to train, and you can get basically spot-on results from just 20-30 training images.
Let's say you want to train an AI in a specific artist's style, or to mimic a specific character, or even a specific behavior (boobjobs, for example.)

Honestly, best place to look is all the OPs resources on /g/, /h/, and the other boards.
Look for /sdg/ on /g/ and /hdg/ on /h/. You will get good information there for sure!

You can definitely re-use the same girl by making a LORA of her, or just re-using the same seed and tweaking qualities without changing too many parameters.
Having the same seed and a very exact phrase will basically net you a very similar girl each time, as long as you use a few "key" parameters.

You can use image-to-image conversion too, and just poorly draw a girl that the algorithm will turn into a supermodel.
Interesting stuff.
Sorry for the delay again, brother.

[Anonymous No.53848420]

I think you’re retarded going into it this early with seriousness, but that’s just me. The fact you haven’t monetized it yet shows inadequacy.

[Anonymous No.53848475]

>>53848420
You may have a point, but I raise you another - if by early you refer to Stable Diffusion's readiness for market adoption as early, surely that's what they say about every fresh niche, not yet saturated with bullshit.
People said the same about crypto in 2015.

If by early you're referring to my ineptitude, well, that may be so, but better an early start than a late start, and certainly better than no start at all.
As for why I haven't monetized it yet, well, that itself contradicts what you said - either it's too early to start, or I'm inadequate for not having monetized it yet.

But, I appreciate the criticism, anon.
I will keep it in mind.

[Anonymous No.53848654]

>>53848475
https://voca.ro/14qtN0fq4L0g
Unless you have something other than what a retail user can pull off, I think you are setting yourself up for failure.

[Anonymous No.53848660]

>>53848654
My ID changed.

[Anonymous No.53848764]

>>53848660
I know the feel, anon.
I do have a few things though - I will also remind you that just hosting Stable Diffusion is probably beyond what 99% of normies are capable of.
You are overestimating the abilities of the average media consoomer - most of these people couldn't download Microsoft Office, and I can back this with the fact that I worked in IT for 4 years.

However, I'm planning to specialize in on-demand hypernetworks / LoRA training.

[Anonymous No.53848846]

>>53848654
To respond to your message, anon.
I appreciate your genuinely in-depth elaboration and criticism.
I think I will listen to it a few times over the next day to fully absorb everything you've said.

I disagree though with your idea of mental masturbation - in a funny way, you do remind me a lot of my father - he says the same, always has said it. I greatly respect my father.
Still, while I *also* respect the appreciation for pragmatism, I find the fixation on "one-track" action, without a moment spent to ponder or consider something else. In fact, that's been the undoing of many autists.
Training oneself to be a machine has failed time and time again - sure, they can accomplish greatness in their one ambition, but their life is lacking in every other aspect.

I will elaborate further - we mustn't lose sight of where we are.
It's a silly thread created on a Mongolian basket-weaving forum.
Procrastination and recreation / conversation are two different things, and I find that being a little silly can go a long way to opening up opportunities.

To answer your question - I have been programming for 5 years.
I specialize in C++, Solidity, and Java. Been web developing another 3 years (HTML/CSS/JS stack.)
You are right though - I do have very little experience in ML or Python, as I said, only a few months, and that was on-and-off.

You are right about one more thing - that I should act. I will point out another thing though - I have never delegated work in my life. Of course, I have not managed to address your message in its entirety within the confines of this post, but it is late, and I best sleep on it first. In all honesty, I may be rambling now, so that is where I will end it and say this - cheers.

P.S. You are an interesting chap.
[e-mail]
I'd be happy to talk some more in the future.

[Anonymous No.53848904]

>>53848764
Well as long as you have the ability to do so and do not intend to larp or delude yourself then I don’t have an issue. I would say that “hyper networks” are just adjusters, so unless you’re receiving fees it again seems rather silly. If I was to pursue AI I would immediately work towards building an application that was consistent with bone rigging, a usable 3D space for imports, and have a front end where you can draw from presets that are layered like Photoshop, (maybe in the vein like LORA you keep specifying or not I don’t know). You could adjust percentiles of character distribution based on camera angle, and create a final filter processing mode at the end of the frame for aesthetic. I would start selling licenses to the animation industry (precluding any legal issues that will occur). This is already something rather simple to do for a CS graduate within the confines of the current released public builds, and would generate millions in licensing and untold IP which can further be sold if you’re even a tad creative. The problem arises on whether you actually have any real capabilities and ideas that extend beyond middler ability. Lora and hyper networks seem like you’re only using them in a semantic usage, rather than real world application at the moment. I don’t actually like using the term midwit to refer to people, but it was done so in the vocaroo to highlight the fact you may not have the talent required to actually monetize anything meaningful aside from retail limits. Retail does not just consider boomers and Stacy, it implies engineers and workers who are graduates with CS and STEM degrees. That is retail. So where do you compare to your competition?

[Anonymous No.53848919]

>>53848904
Ah I did not see your subsequent reply, nice! As long as you can develop apps and understand math and abstracts I wish you all the good luck.

[Anonymous No.53849098]

>>53848904
*seems rather silly unless you are building something specific with it. Another app idea I had was using human labor on a mobile app to adjust AI output collectively, users highlight errors in the picture with marquee tool or are presented with 1 or 2, A or B scenarios of what looks better, which then feeds back into your client(s) own AI model to correct issue with generation missteps of the model or human preference to whatever is more aesthetically pleasant. Revision or revert could be made at any time by keeping user data in iterations and versions alongside manual moderating, so if the client did not like where his model was being pushed this could be reverted. Users could be paid to do this in terms of pennies or rather, be able to use the clients models for their own generative purposes via the app. Personally I don’t see much money in this idea due to how much progress is being made at the rate it is going, but to data mine AI preference from users could be useful.