Zepph
Registered Member
- Joined
- Apr 22, 2020
- Messages
- 87
- Reaction score
- 59
Some background about me
Two years ago I wanted to make enough money to live comfortably. I made an Instagram account in the cat niche and just reuploaded scraped images from other profiles. I was able to go from 0 to 10k followers in half a year without investments, just using the follow x follow method and posting daily.
A year and a half ago I got interested in OpenAI, a company whose sole purpose was to design innovative AI models. When they released GPT-3 I was incredibly hyped. I talked to everyone around me about it, nobody could make me stop, but very few people cared about it. I saw from that moment onwards the revolutionary consequences artificial intelligence would have in the following years.
A year and three months ago I discovered Dall-E, a fun but promising project from OpenAI to make image generations from text prompts. I also talked about it non-stop. It truly was fascinating to read the papers and try to understand whatever the hell was going on behind its algorithms.
A year ago I was already waiting for the public release of Dall-E 2. I was so excited that I even tried to get into the first testers just for fun. Out of nowhere, OpenAI gave me early access to such tool and it was an incredible experience to use such state of the art technology.
Half a year algo I noticed that Dall-E 2 was not anymore the state of the art, the competitors were now Stable Diffusion and Midjourney. I didn't have any interest in Stable Diffusion, but I did with Midjourney. I tried it out a little, but it didn't seem that amazing at first glance.
Two months ago I got to know about ChatGPT, a fine-tuned model of GPT considered by many GPT-3.5, and what I had seen coming from a year and a half ago became even more of a reality than a dream. Even since I have used ChatGPT in quite a lot of tasks, and I have made money using it too, but not enough for what I want to achieve. At the same time, I tried to use Midjourney again, just for fun.
In the past, once I achieved success with my Instagram account I didn't know what to do with it and just left it to die... Even if I failed in the past I learnt a valuable lesson, I need to have a plan to actually make the profit I desire.
On the 17th of January of 2023, I decided to do something, a new project that could take advantage of my Instagram knowledge and my enthusiasm for AI.
The plan
In short, I'll make an Instagram account. I'll use the follow x follow method to grow it and then I'll send merch based on the images I upload. Pretty standard right? Where's the twist then? Well, I think you have already guessed it if you read my background. These images will be AI generated. For the merch I'll use a webpage named Printful and then I'll share it once I have a decent amount of followers.
The first problem
Currently, I don't have much money, here in my country the wages are terrible so the only money I make is from freelance jobs on programming. I have a very limited budget. My first option was to research about Stable Diffusion, which is an open-source alternative for AI generation. It's pretty damn good to be honest, specially their fine-tuned models like Anything V3. Quite viable, but I don't have a GPU and configuring the Collab notebooks can be a bit of a hassle, even if they're writing in my main programming language, Python. Plus, the style doesn't convince me at all when compared to V4 Midjourney, the state-of-the-art technology on image generation, which is about $30 per month for unlimited generations. And I'll be honest, $30 is a ton money for me at the moment, however it's a risk I can take without going bankrupt, and it's my best option as of right now.
With that money spent, it was time to learn how to properly use Midjourney. Yes, it's as simple as just writing whatever the hell I want and it'll just spit out some amazing artwork, but I've got a feeling that I can make it even better with some experimenting and reading.
After a few days of experimenting, reading and tweaking its settings, I was at a comfortable place with manipulating Midjourney to my desires.
The second problem
By this time I had already made a new e-mail account and already uploaded some Stable Diffusion generated images using a cloud-based service for faster generations than using collab or my CPU (I used its trial). I was already using my follow x follow method and I already had about 10 followers and about 20 reactions to my posts (only two posts lol), but I knew that if I wanted to sell something I had to have an art niche to focus. After thinking for a while, I thought of a niche related to the cat niche I had in my past project and I started generated images.
The third problem
Now, the problem was that I ran out of prompt ideas very very quickly. I guess knowing how to describe artwork doesn't give you the creativity of an artist, what a surprise right?
Nonetheless, I already knew that I was no artist. I am an AI enthusiast, and I also know how to use ChatGPT with fluency, so I just talked him into giving me ideas for drawings in the style I wanted. I've now automated the entire creative process and the only job I've got to do is to generate the images, select those who I think look good and then upload them.
The fourth problem
I started to notice that a lot of the generations were almost perfect, but not quite. Midjourney has an option to make variations of a certain image, and iterating through the process for a few times usually fixed the slight errors, but it takes too much time for my taste. The solution I found was pretty simple, I'll just edit out the problems manually, which is what I've been doing these days and it has worked seamlessly.
The fifth and final problem
This is the least and most important problem as of right now. I only have 17 days left to use the unlimited generations from Midjourney, but I'm having trouble on finding motivation to keep going. I have automated the entire process but now it's just boring. My solution to this is to keep on updating my progress here so I feel obliged (in a good way) to keep on going. My initial goal was to generate 365 images in this month to post one daily, I've already generated about 50 of them, but as of right now that would equal 315 left to go, which would equal to 315/17=18 images per day, quite possible but I don't know if I can do it. So my idea was to half those 365 images (315 now) to about 182, posting on my Instagram every two days instead of daily, which means about 9 images per day. Still hard but I think I can do it.
Progress so far
Since the 17 of January I've been able to get about 60 followers and an average of 30 reactions and 1-2 comments per post. I think I'm doing good so far. I hope I can keep on going.
Two years ago I wanted to make enough money to live comfortably. I made an Instagram account in the cat niche and just reuploaded scraped images from other profiles. I was able to go from 0 to 10k followers in half a year without investments, just using the follow x follow method and posting daily.
A year and a half ago I got interested in OpenAI, a company whose sole purpose was to design innovative AI models. When they released GPT-3 I was incredibly hyped. I talked to everyone around me about it, nobody could make me stop, but very few people cared about it. I saw from that moment onwards the revolutionary consequences artificial intelligence would have in the following years.
A year and three months ago I discovered Dall-E, a fun but promising project from OpenAI to make image generations from text prompts. I also talked about it non-stop. It truly was fascinating to read the papers and try to understand whatever the hell was going on behind its algorithms.
A year ago I was already waiting for the public release of Dall-E 2. I was so excited that I even tried to get into the first testers just for fun. Out of nowhere, OpenAI gave me early access to such tool and it was an incredible experience to use such state of the art technology.
Half a year algo I noticed that Dall-E 2 was not anymore the state of the art, the competitors were now Stable Diffusion and Midjourney. I didn't have any interest in Stable Diffusion, but I did with Midjourney. I tried it out a little, but it didn't seem that amazing at first glance.
Two months ago I got to know about ChatGPT, a fine-tuned model of GPT considered by many GPT-3.5, and what I had seen coming from a year and a half ago became even more of a reality than a dream. Even since I have used ChatGPT in quite a lot of tasks, and I have made money using it too, but not enough for what I want to achieve. At the same time, I tried to use Midjourney again, just for fun.
In the past, once I achieved success with my Instagram account I didn't know what to do with it and just left it to die... Even if I failed in the past I learnt a valuable lesson, I need to have a plan to actually make the profit I desire.
On the 17th of January of 2023, I decided to do something, a new project that could take advantage of my Instagram knowledge and my enthusiasm for AI.
The plan
In short, I'll make an Instagram account. I'll use the follow x follow method to grow it and then I'll send merch based on the images I upload. Pretty standard right? Where's the twist then? Well, I think you have already guessed it if you read my background. These images will be AI generated. For the merch I'll use a webpage named Printful and then I'll share it once I have a decent amount of followers.
The first problem
Currently, I don't have much money, here in my country the wages are terrible so the only money I make is from freelance jobs on programming. I have a very limited budget. My first option was to research about Stable Diffusion, which is an open-source alternative for AI generation. It's pretty damn good to be honest, specially their fine-tuned models like Anything V3. Quite viable, but I don't have a GPU and configuring the Collab notebooks can be a bit of a hassle, even if they're writing in my main programming language, Python. Plus, the style doesn't convince me at all when compared to V4 Midjourney, the state-of-the-art technology on image generation, which is about $30 per month for unlimited generations. And I'll be honest, $30 is a ton money for me at the moment, however it's a risk I can take without going bankrupt, and it's my best option as of right now.
With that money spent, it was time to learn how to properly use Midjourney. Yes, it's as simple as just writing whatever the hell I want and it'll just spit out some amazing artwork, but I've got a feeling that I can make it even better with some experimenting and reading.
After a few days of experimenting, reading and tweaking its settings, I was at a comfortable place with manipulating Midjourney to my desires.
The second problem
By this time I had already made a new e-mail account and already uploaded some Stable Diffusion generated images using a cloud-based service for faster generations than using collab or my CPU (I used its trial). I was already using my follow x follow method and I already had about 10 followers and about 20 reactions to my posts (only two posts lol), but I knew that if I wanted to sell something I had to have an art niche to focus. After thinking for a while, I thought of a niche related to the cat niche I had in my past project and I started generated images.
The third problem
Now, the problem was that I ran out of prompt ideas very very quickly. I guess knowing how to describe artwork doesn't give you the creativity of an artist, what a surprise right?
Nonetheless, I already knew that I was no artist. I am an AI enthusiast, and I also know how to use ChatGPT with fluency, so I just talked him into giving me ideas for drawings in the style I wanted. I've now automated the entire creative process and the only job I've got to do is to generate the images, select those who I think look good and then upload them.
The fourth problem
I started to notice that a lot of the generations were almost perfect, but not quite. Midjourney has an option to make variations of a certain image, and iterating through the process for a few times usually fixed the slight errors, but it takes too much time for my taste. The solution I found was pretty simple, I'll just edit out the problems manually, which is what I've been doing these days and it has worked seamlessly.
The fifth and final problem
This is the least and most important problem as of right now. I only have 17 days left to use the unlimited generations from Midjourney, but I'm having trouble on finding motivation to keep going. I have automated the entire process but now it's just boring. My solution to this is to keep on updating my progress here so I feel obliged (in a good way) to keep on going. My initial goal was to generate 365 images in this month to post one daily, I've already generated about 50 of them, but as of right now that would equal 315 left to go, which would equal to 315/17=18 images per day, quite possible but I don't know if I can do it. So my idea was to half those 365 images (315 now) to about 182, posting on my Instagram every two days instead of daily, which means about 9 images per day. Still hard but I think I can do it.
Progress so far
Since the 17 of January I've been able to get about 60 followers and an average of 30 reactions and 1-2 comments per post. I think I'm doing good so far. I hope I can keep on going.
.
