Before we start the article, let's test your ability to recognize AI. Which of the images below are photos of real people? Which ones are generated by AI?
To reveal the answer to the question.The images above are all AI generated!
This is from a chat record that has been widely circulated in WeChat groups lately, and I don't know if you've seen it:
A big brother draws photos of beautiful women in AI.Batch operation of more than 1000 by one personLittle Red Bookaccount, dominating the front page of Little Red Book for the keyword "beauty".
Almost all of the accounts in the picture below were messed up by him alone!
Generated with AIPictures of beautiful womenNot only were they not recognized by the platform, but they also gave push streams!It really is ridiculously strong.
Awesome, right?
But AI making beautiful women isn't a new project.
There have been many AI beauty numbers popping up on major self-media platforms, but because whether it's a beauty made by MJ or SD, there will be a more obvious AI flavor, such as the wrong number of fingers in the AI drawing, or a strong sense of peeling that is not real.
There were also some dedicated AI beauty-fighting bloggers as a result, so it was short-lived after platforms labeled this type of content 'AI-generated'.
And not only were the images produced by this big brother not labeled 'AI-generated', they were even pushed into the stream, completely solving the previous problem. Why?
It's not really that Big Brother's production skills are so great, but the new Flux models came out with a huge increase in capability compared to the previous drawing tools.
Today we're going to take a peek at this Big Brother operation, including:
I. Realization logic
II. Production methods
I. Realization logic
1. Up the powder to sell accounts
According to Big Brother's own description, he is currently cashing in mainly by raising accounts and selling them.
We know that although beautiful bloggers are quick to increase their fan base, there is no commercial value in taking ads for this kind of generalized fan base.
However, the platform to open the window of the goods hanging goods is required to have a certain number of fans, for example, Xiaohongshu Dandelion threshold is 1000 powder. So merchants need a large number of accounts that can hang goods, giving rise to such a path to cash.
2. Selling peripherals to LSPs
This type of account focus on the fans are basically male LSP, through the account diversion to the private domain, sell some LSP like peripheral products (as for which products you brainstorm, do not understand to go to the beauty of the live room to add a WeChat to see), LSP their spending power is still relatively strong.
3. Beauty IP
The hardest and highest ceiling:
Relying on improved AI drawing capabilities, you can maintain consistency in your characters and easily create scenes of all types.
"Fixed Characters" + "Vertical Content Scene" = "Beauty Personal IP".
Similarly, Aitana Lopez, who was an ins sensation before, is said to be earning more than $100,000 a month, and lingerie giant Victoria's Secret has partnered with her.
Throwing a brick to attract jade, for example, such as the recent relatively hot girls' cycling, it can be coupled with vertical content output, to create a bicycle riding beauty, such an account is very valuable, look at the comment area of the benchmark account, there are many bicycles, clothes, equipment, the demand for bandwagon.
Of course this one requires the highest level of AI mapping skills.
II. Production methods
Generate it with Flux. There are many ways to call Flux, whether it's Webui or Comfyui.
The simpler one is to use the domestic Liblib (https://www.liblib.art/), which has a very large number of well-trained Flux LORAs, and 300 points per day for free use, which is perfectly adequate for newcomers to graph generation.
But the premise is the need to have a better foundation in AI drawing, this topic is more complex, not to expand on it, interested in tutorials you can find to see, you can also look for me to be more reliable tutorials suggested.
Let's try one of the easiest ones first, one that any white person can follow:
1. Open Liblib, search and select an Asian face LORA for Flux:
Take this one, for example:
2. Open the Flux basic algorithm that has been set up on it, Lora has also selected the Asian portrait that has just been selected.
So just type in the cue word and just, click on the one-click raw image.
TIPS: If you don't know how to write a cue word, you can find a counterpart photo you want to emulate, open the WD1.4 tagger, and upload it directly for the AI to reverse-engineer one.
Put up a couple random cards I drew to see the effect:
Well, that's it for simple image making. If you want to get advanced, you need to systematically learn more AI PaintingTip up.
However, it does not necessarily have to be technology, the key is how to realize the latter, think clearly about this issue, matching the appropriate level of technology can be.
We can also think for ourselves, how else can such technology be realized?