Honestly, I used to think that generating AI images was just a gimmick.
Back in March last year, I spent over 3,000 yuan on every "super prompt package" I could find. I tried everything—Midjourney, Stable Diffusion, and various local platforms. And you know what? Most of them just didn’t work.
That “one-click masterpiece”? Either the fingers looked deformed or the images had a cheesy plastic vibe. I found myself staring blankly at the screen, feeling like I was just paying tuition.
But then, a turning point came.
I’m not a photographer, but I stumbled upon something fascinating on Xiaohongshu: someone was selling a 1-yuan AI photo package that sold over 2,000 copies. What could you possibly earn from 1 yuan? I clicked in, and to my surprise, the comments were flooded with requests like “Can I get baby photo prompts?” and “How do I adjust for ancient-style wedding photos?”
In that moment, it clicked for me: users weren’t just buying prompts; they were buying the certainty of “if I input this, I’ll get a great image.”
Here’s the first surprising insight:
the essence of the AI image business is selling certainty, not just prompts.
So I changed my approach. Instead of piling up 50 words for prompts, I started testing combinations of 20 words: “subject + style + atmosphere + what to avoid.” I focused on “what to avoid”—no blurriness, no plastic feel, no deformed fingers.
After testing about 30 sets, I found three stable “formulas” that consistently produced images scoring 80 out of 100.
Then I did something kind of silly but super effective: I took one of the “ancient-style wedding photo” prompts, priced it at 9.9 yuan, and shared it on my social networks.
Within 24 hours, 17 people made a purchase. I made 168.3 yuan. Not much, but enough to keep me awake from excitement.
Here comes the second surprising insight:
the real money isn’t in hoarding prompts, but in continuous sharing.
I began releasing a new style prompt package each week, posting example images on Xiaohongshu and Douyin. Whether they were polished images or directly generated ones—even keeping some AI “flaws”—it didn’t matter.
Soon, people started asking in the comments: “Can this be used commercially?” “Can I train it on my own face?”
That’s when the business opportunities kicked in.
So, I launched a second product: a 599 yuan “AI Writing Practical Course.” I didn’t teach complicated parameters, just three things:
1) How to determine if an image can go viral;
2) How to consistently generate images using 20 words;
3) How to post images on Xiaohongshu without getting shadowbanned.
In the first month, I sold 23 copies, generating six-figure revenue.
But here’s the cold, hard truth: there’s a big pitfall here.
Many people get a little too excited at this stage, rushing to recruit agents and launching overpriced courses. I’ve seen someone sell over 100 copies of a 2,999 yuan course, but their after-sales support couldn't keep up, and two months later, their reputation crashed, and their account was ruined.
My approach was different.
I selected the first batch of students who bought the course and actually achieved results, and I upgraded five of them to “partners” for free. They received a higher commission but had strict requirements: they had to produce three new prompt sets weekly and publish examples on their platforms.
Now, I have over 30 such partners.
This is the true barrier: it’s not about how good your prompts are, but how many people you can get to help you share images and reach wider audiences.
Last month, I did some calculations: prompt package revenue only accounted for 30%, the course for 40%, and partner commissions for 30%. But the most important part? My example images reach at least 500,000 people daily across Douyin, Xiaohongshu, and video platforms.
As users scroll, they see our images. The more they see, the more they think, “Wow, this AI photography really looks good.”
That's how aesthetic authority is built.
But I have to be honest: there’s no permanent moat in this industry. AI tools are evolving every day, and a prompt that works today could be obsolete tomorrow.
The only real moat is your speed.
You have to test new styles faster, cover new platforms quicker, and turn user feedback into product iterations swiftly. Speed is everything.
One last heartfelt piece of advice:
If you’re still collecting prompt files, hoarding 10GB without ever posting a single image, you’re wasting your time.
Starting today, pick a style that resonates with you (be it ancient, cyberpunk, or baby photography), generate 9 images, price them at 19.9 yuan, and put them out there to see what happens.
You’ll either make money or get real user feedback. Either outcome is better than just stockpiling prompts.
Stop hoarding—start sharing. If you don’t post today, someone else will use your style to make money tomorrow.
Top comments (0)