How I monetized and sold an abandoned side project for $20,000

The Origin Story

I've always wanted to be an entrepreneur but never succeeded. I often had business ideas, made a landing page, and told my buddies. Never got customers.

In April 2021, I decided to try again with a new strategy. I noticed that I had trouble acquiring an initial set of customers, so I wanted to start by acquiring a product that had a small user base that I could grow.

I found a SaaS marketplace called MicroAcquire.com where you could buy and sell SaaS products. I liked Shareit.video, an online Loom-like screen recorder.

Shareit.video didn't generate revenue, but 50 people visited daily to record screencasts.

Purchasing a Failed Side Project

I eventually bought Shareit.video for $12,000 from its owner.

$12,000 was probably too much for a website without revenue or registered users.

I thought time was most important. I could have recreated the website, but it would take months. $12,000 would give me an organized code base and a working product with a few users to monetize.

I considered buying a screen recording website and trying to grow it versus buying a new car or investing in crypto with the $12K.

Buying the website would make me a real entrepreneur, which I wanted more than anything.

Putting down so much money would force me to commit to the project and prevent me from quitting too soon.

A Year of Development

I rebranded the website to be called RecordJoy and worked on it with my cousin for about a year. Within a year, we made $5000 and had 3000 users.

We spent $3500 on ads, hosting, and software to run the business.

AppSumo promoted our $120 Life Time Deal in exchange for 30% of the revenue.

We put RecordJoy on maintenance mode after 6 months because we couldn't find a scalable user acquisition channel.

We improved SEO and redesigned our landing page, but nothing worked.

Despite not being able to grow RecordJoy any further, I had already learned so much from working on the project so I was fine with putting it on maintenance mode. RecordJoy still made $500 a month, which was great lunch money.

Getting Taken Over

One of our customers emailed me asking for some feature requests and I replied that we weren’t going to add any more features in the near future. They asked if we'd sell.

We got on a call with the customer and I asked if he would be interested in buying RecordJoy for 15k. The customer wanted around $8k but would consider it.

Since we were negotiating with one buyer, we put RecordJoy on MicroAcquire to see if there were other offers.

We quickly received 10+ offers. We got 18.5k. There was also about $1000 in AppSumo that we could not withdraw, so we agreed to transfer that over for $600 since about 40% of our sales on AppSumo usually end up being refunded.

Lessons Learned

First, create an acquisition channel

We couldn't discover a scalable acquisition route for RecordJoy. If I had to start another project, I'd develop a robust acquisition channel first. It might be LinkedIn, Medium, or YouTube.

Purchase Power of the Buyer Affects Acquisition Price

Some of the buyers we spoke to were individuals looking to buy side projects, as well as companies looking to launch a new product category. Individual buyers had less budgets than organizations.

Customers of AppSumo vary.

AppSumo customers value lifetime deals and low prices, which may not be a good way to build a business with recurring revenue. Designed for AppSumo users, your product may not connect with other users.

Try to increase acquisition trust

Acquisition often fails. The buyer can go cold feet, cease communicating, or run away with your stuff. Trusting the buyer ensures a smooth asset exchange. First acquisition meeting was unpleasant and price negotiation was tight. In later meetings, we spent the first few minutes trying to get to know the buyer’s motivations and background before jumping into the negotiation, which helped build trust.

Operating expenses can reduce your earnings.

Monitor operating costs. We were really happy when we withdrew the $5000 we made from AppSumo and Stripe until we realized that we had spent $3500 in operating fees. Spend money on software and consultants to help you understand what to build.

Don't overspend on advertising

We invested $1500 on Google Ads but made little money. For a side project, it’s better to focus on organic traffic from SEO rather than paid ads unless you know your ads are going to have a positive ROI.

Price Intelligently put out amazing content on pricing your SaaS product. This blog's link to the whole report is worth reading. Our key takeaways are below.

Don't base prices on the competition. Competitor-based pricing has clear drawbacks. Their pricing approach is yours. Your company offers customers something unique. Otherwise, you wouldn't create it. This strategy is static, therefore you can't add value by raising prices without outpricing competitors. Look, but don't touch is the competitor-based moral. You want to know your competitors' prices so you're in the same ballpark, but they shouldn't guide your selections. Competitor-based pricing also drives down prices.

Value-based pricing wins. This is customer-based pricing. Value-based pricing looks outward, not inward or laterally at competitors. Your clients are the best source of pricing information. By valuing customer comments, you're focusing on buyers. They'll decide if your pricing and packaging are right. In addition to asking consumers about cost savings or revenue increases, look at data like number of users, usage per user, etc.

Value-based pricing increases prices. As you learn more about the client and your worth, you'll know when and how much to boost rates. Every 6 months, examine pricing.

Cloning top customers. You clone your consumers by learning as much as you can about them and then reaching out to comparable people or organizations. You can't accomplish this without knowing your customers. Segmenting and reproducing them requires as much detail as feasible. Offer pricing plans and feature packages for 4 personas. The top plan should state Contact Us. Your highest-value customers want more advice and support.

Question your 4 personas. What's the one item you can't live without? Which integrations matter most? Do you do analytics? Is support important or does your company self-solve? What's too cheap? What's too expensive?

Not everyone likes per-user pricing. SaaS organizations often default to per-user analytics. About 80% of companies utilizing per-user pricing should use an alternative value metric because their goods don't give more value with more users, so charging for them doesn't make sense.

At least 3:1 LTV/CAC. Break even on the customer within 2 years, and LTV to CAC is greater than 3:1. Because customer acquisition costs are paid upfront but SaaS revenues accrue over time, SaaS companies face an early financial shortfall while paying back the CAC.

ROI should be >20:1. Indeed. Ensure the customer's ROI is 20x the product's cost. Microsoft Office costs $80 a year, but consumers would pay much more to maintain it.

A/B Testing. A/B testing is guessing. When your pricing page varies based on assumptions, you'll upset customers. You don't have enough customers anyway. A/B testing optimizes landing pages, design decisions, and other site features when you know the problem but not pricing.

Don't discount. It cheapens the product, makes it permanent, and increases churn. By discounting, you're ruining your pricing analysis.

And tips for anyone trying.

We see the same story every few years.

BREAKING NEWS: [Platform X] launched a social network. With Facebook's reputation down, the new startup bets millions will switch.

Despite the excitement surrounding each new platform (Diaspora, Ello, Path, MeWe, Minds, Vero, etc.), no major exodus occurred.

Snapchat and TikTok attracted teens with fresh experiences (ephemeral messaging and rapid-fire videos). These features aren't Facebook, even if Facebook replicated them.

Facebook's core is simple: you publish items (typically text/images) and your friends (generally people you know IRL) can discuss them.

It's cool. Sometimes I don't want to, but sh*t. I like it.

Because, well, I like many folks I've met. I enjoy keeping in touch with them and their banter.

I dislike Facebook's corporation. I've been cautiously optimistic whenever a Facebook-killer surfaced.

None succeeded.

Why? Two causes, I think:

People couldn't switch quickly enough, which is reason #1

Your buddies make a social network social.

Facebook started in self-contained communities (college campuses) then grew outward. But a new platform can't.

If we're expected to leave Facebook, we want to know that most of our friends will too.

Most Facebook-killers had bottlenecks. You have to waitlist or jump through hoops (e.g. setting up a server).

Same outcome. Upload. Chirp.

After a week or two of silence, individuals returned to Facebook.

Reason #2: The fundamental experience was different.

Even when many of our friends joined in the first few weeks, it wasn't the same.

There were missing features or a different UX.

Want to reply with a meme? No photos in comments yet. (Trying!)

Want to tag a friend? Nope, sorry. 2019!

Want your friends to see your post? You must post to all your friends' servers. Good luck!

It's difficult to introduce a platform with 100% of the same features as one that's been there for 20 years, yet customers want a core experience.

If you can't, they'll depart.

The causes that led to the causes

Having worked on software teams for 14+ years, I'm not surprised by these challenges. They are a natural development of a few tech sector meta-problems:

Lean startup methodology

Silicon Valley worships lean startup. It's a way of developing software that involves testing a stripped-down version with a limited number of people before selecting what to build.

Billion people use Facebook's functions. They aren't tested. It must work right away*

*This may seem weird to software people, but it's how non-software works! You can't sell a car without wheels.

2. Creativity

Startup entrepreneurs build new things, not copies. I understand. Reinventing the wheel is boring.

We know what works. Different experiences raise adoption friction. Once millions have transferred, more features (and a friendlier UX) can be implemented.

3. Cost scaling

True. Building a product that can sustain hundreds of millions of users in weeks is expensive and complex.

Your lifeboats must have the same capacity as the ship you're evacuating. It's required.

4. Pure ideologies

People who work on Facebook-alternatives are (understandably) critical of Facebook.

They build an open-source, fully-distributed, data-portable, interface-customizable, offline-capable, censorship-proof platform.

Prioritizing these aims can prevent replicating the straightforward experience users expect. Github, not Facebook, is for techies only.

What about the business plan, though?

Facebook-killer attempts have followed three models.

1. Utilize VC funding to increase your user base, then monetize them later. (If you do this, you won't kill Facebook; instead, Facebook will become you.)

2. Users must pay to utilize it. (This causes a huge bottleneck and slows the required quick expansion, preventing it from seeming like a true social network.)

3. Make it a volunteer-run, open-source endeavor that is free. (This typically denotes that something is cumbersome, difficult to operate, and is only for techies.)

Wikipedia is a fourth way.

Wikipedia is one of the most popular websites and a charity. No ads. Donations support them.

A Facebook-killer managed by a good team may gather millions (from affluent contributors and the crowd) for their initial phase of development. Then it might sustain on regular donations, ethical transactions (e.g. fees on commerce, business sites, etc.), and government grants/subsidies (since it would essentially be a public utility).

When you're not aiming to make investors rich, it's remarkable how little money you need.

If you want to build a Facebook competitor, follow these tips:

1. Drop the lean startup philosophy. Wait until you have a finished product before launching. Build it, thoroughly test it for bugs, and then release it.

2. Delay innovating. Wait till millions of people have switched before introducing your great new features. Make it nearly identical for now.

3. Spend money climbing. Make sure that guests can arrive as soon as they are invited. Never keep them waiting. Make things easy for them.

4. Make it accessible to all. Even if doing so renders it less philosophically pure, it shouldn't require technical expertise to utilize.

5. Constitute a nonprofit. Additionally, develop community ownership structures. Profit maximization is not the only strategy for preserving valued assets.

Last thoughts

Nobody has killed Facebook, but Facebook is killing itself.

The startup is burying the newsfeed to become a TikTok clone. Meta itself seems to be ditching the platform for the metaverse.

I wish I was happy, but I'm not. I miss (understandably) removed friends' postings and remarks. It could be a ghost town in a few years. My dance moves aren't TikTok-worthy.

Who will lead? It's time to develop a social network for the people.

Greetings if you're working on it. I'm not a company founder, but I like to help hard-working folks.

ZERO GRACE/ZERO MALICE

Amazon's purchase of One Medical could speed up American healthcare

The U.S. healthcare industry is a 7-ton seal bleeding at sea. Predators are circling. Unearned margin: price increases relative to inflation without quality improvements. Amazon is the 11-foot megalodon with 7-inch teeth. Amazon is no longer circling... but attacking.

In 2020 dollars, per capita U.S. healthcare spending increased from $2,968 in 1980 to $12,531. The result is a massive industry with 13% of the nation's workers and a fifth of GDP.

Doctor No

In 40 years, healthcare has made progress. From 73.7 in 1980 to 78.8 in 2019, life expectancy rose (before Covid knocked it back down a bit). Pharmacological therapies have revolutionized, and genetic research is paying off. The financial return, improvement split by cost increases, is terrible. No country has expense rises like the U.S., and no one spends as much per capita as we do. Developed countries have longer life expectancies, healthier populations, and less economic hardship.

Two-thirds of U.S. personal bankruptcies are due to medical expenses and/or missed work. Mom or Dad getting cancer could bankrupt many middle-class American families. 40% of American adults delayed or skipped needed care due to cost. Every healthcare improvement seems to have a downside. Same pharmacological revolution that helped millions caused opioid epidemic. Our results are poor in many areas: The U.S. has a high infant mortality rate.

Healthcare is the second-worst retail industry in the country. Gas stations are #1. Imagine walking into a Best Buy to buy a TV and a Blue Shirt associate requests you fill out the same 14 pages of paperwork you filled out yesterday. Then you wait in a crowded room until they call you, 20 minutes after the scheduled appointment you were asked to arrive early for, to see the one person in the store who can talk to you about TVs, who has 10 minutes for you. The average emergency room wait time in New York is 6 hours and 10 minutes.

If it's bad for the customer, it's worse for the business. Physicians spend 27% of their time helping patients; 49% on EHRs. Documentation, order entry, billing, and inbox management. Spend a decade getting an M.D., then become a bureaucrat.

No industry better illustrates scale diseconomies. If we got the same return on healthcare spending as other countries, we'd all live to 100. We could spend less, live longer and healthier, and pay off the national debt in 15 years. U.S. healthcare is the worst ever.

What now? Competition is at the heart of capitalism, the worst system of its kind.

Priority Time

Amazon is buying One Medical for $3.9 billion. I think this deal will liberate society. Two years in, I think One Medical is great. When I got Covid, I pressed the One Medical symbol on my phone; a nurse practitioner prescribed Paxlovid and told me which pharmacies had it in stock.

Amazon enables the company's vision. One Medical's stock is down to $10 from $40 at the start of 2021. Last year, it lost $250 million and needs cash (Amazon has $60 billion). ONEM must grow. The service has 736,000 members. Half of U.S. households have Amazon Prime. Finally, delivery. One Medical is a digital health/physical office hybrid, but you must pick up medication at the pharmacy. Upgrade your Paxlovid delivery time after a remote consultation. Amazon's core competency means it'll happen. Healthcare speed and convenience will feel alien.

It's been a long, winding road to disruption. Amazon, JPMorgan, and Berkshire Hathaway formed Haven four years ago to provide better healthcare for their 1.5 million employees. It rocked healthcare stocks the morning of the press release, but folded in 2021.

Amazon Care is an employee-focused service. Home-delivered virtual health services and nurses. It's doing well, expanding nationwide, and providing healthcare for other companies. Hilton is Amazon Care's biggest customer. The acquisition of One Medical will bring 66 million Prime households capital, domain expertise, and billing infrastructure. Imagine:

"Alexa, I'm hot and my back hurts."

"Connecting you to a Prime doctor now."

Want to vs. Have to

I predicted Amazon entering healthcare years ago. Why? For the same reason Apple is getting into auto. Amazon's P/E is 56, double Walmart's. The corporation must add $250 billion in revenue over the next five years to retain its share price. White-label clothes or smart home products won't generate as much revenue. It must enter a huge market without scale, operational competence, and data skills.

Current Situation

Healthcare reform benefits both consumers and investors. In 2015, healthcare services had S&P 500-average multiples. The market is losing faith in public healthcare businesses' growth. Healthcare services have lower EV/EBITDA multiples than the S&P 500.

Amazon isn't the only prey-hunter. Walmart and Alibaba are starting pharmacies. Uber is developing medical transportation. Private markets invested $29 billion in telehealth last year, up 95% from 2020.

The pandemic accelerated telehealth, the immediate unlock. After the first positive Covid case in the U.S., services that had to be delivered in person shifted to Zoom... We lived. We grew. Video house calls continued after in-person visits were allowed. McKinsey estimates telehealth visits are 38 times pre-pandemic levels. Doctors adopted the technology, regulators loosened restrictions, and patients saved time. We're far from remote surgery, but many patient visits are unnecessary. A study of 40 million patients during lockdown found that for chronic disease patients, online visits didn't affect outcomes. This method of care will only improve.

Amazon's disruption will be significant and will inspire a flood of capital, startups, and consumer brands. Mark Cuban launched a pharmacy that eliminates middlemen in January. Outcome? A 90-day supply of acid-reflux medication costs $17. Medicare could have saved $3.6 billion by buying generic drugs from Cuban's pharmacy. Other apex predators will look at different limbs of the carcass for food. Nike could enter healthcare via orthopedics, acupuncture, and chiropractic. LVMH, L'Oréal, and Estée Lauder may launch global plastic surgery brands. Hilton and Four Seasons may open hospitals. Lennar and Pulte could build "Active Living" communities that Nana would leave feet first, avoiding the expense and tragedy of dying among strangers.

Risks

Privacy matters: HIV status is different from credit card and billing address. Most customers (60%) feel fine sharing personal health data via virtual technologies, though. Unavoidable. 85% of doctors believe data-sharing and interoperability will become the norm. Amazon is the most trusted tech company for handling personal data. Not Meta: Amazon.

What about antitrust, then?

Amazon should be required to spin off AWS and/or Amazon Fulfillment and banned from promoting its own products. It should be allowed to acquire hospitals. One Medical's $3.9 billion acquisition is a drop in the bucket compared to UnitedHealth's $498 billion market valuation.

Antitrust enforcement shouldn't assume some people/firms are good/bad. It should recognize that competition is good and focus on making markets more competitive in each deal. The FTC should force asset divestitures in e-commerce, digital marketing, and social media. These companies can also promote competition in a social ill.

U.S. healthcare makes us fat, depressed, and broke. Competition has produced massive value and prosperity across most of our economy.

Dear Amazon … bring it.

Deep learning-based generative art is being researched. As usual, self-learning is better. Some models, like OpenAI's DALL-E 2, require registration and can only be used online, but others can be used locally, which is usually more enjoyable for curious users. I'll demonstrate the Stable Diffusion model's operation on a standard PC.

Let’s get started.

What It Does

Stable Diffusion uses numerous components:

• A generative model trained to produce images is called a diffusion model. The model is incrementally improving the starting data, which is only random noise. The model has an image, and while it is being trained, the reversed process is being used to add noise to the image. Being able to reverse this procedure and create images from noise is where the true magic is (more details and samples can be found in the paper).

• An internal compressed representation of a latent diffusion model, which may be altered to produce the desired images, is used (more details can be found in the paper). The capacity to fine-tune the generation process is essential because producing pictures at random is not very attractive (as we can see, for instance, in Generative Adversarial Networks).

• A neural network model called CLIP (Contrastive Language-Image Pre-training) is used to translate natural language prompts into vector representations. This model, which was trained on 400,000,000 image-text pairs, enables the transformation of a text prompt into a latent space for the diffusion model in the scenario of stable diffusion (more details in that paper).

This figure shows all data flow:

The weights file size for Stable Diffusion model v1 is 4 GB and v2 is 5 GB, making the model quite huge. The v1 model was trained on 256x256 and 512x512 LAION-5B pictures on a 4,000 GPU cluster using over 150.000 NVIDIA A100 GPU hours. The open-source pre-trained model is helpful for us. And we will.

Install

Before utilizing the Python sources for Stable Diffusion v1 on GitHub, we must install Miniconda (assuming Git and Python are already installed):

wget https://repo.anaconda.com/miniconda/Miniconda3-py39_4.12.0-Linux-x86_64.sh
chmod +x Miniconda3-py39_4.12.0-Linux-x86_64.sh
./Miniconda3-py39_4.12.0-Linux-x86_64.sh
conda update -n base -c defaults conda

Install the source and prepare the environment:

git clone https://github.com/CompVis/stable-diffusion
cd stable-diffusion
conda env create -f environment.yaml
conda activate ldm
pip3 install transformers --upgrade

Download the pre-trained model weights next. HiggingFace has the newest checkpoint sd-v14.ckpt (a download is free but registration is required). Put the file in the project folder and have fun:

python3 scripts/txt2img.py --prompt "hello world" --plms --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

Almost. The installation is complete for happy users of current GPUs with 12 GB or more VRAM. RuntimeError: CUDA out of memory will occur otherwise. Two solutions exist.

Running the optimized version

Try optimizing first. After cloning the repository and enabling the environment (as previously), we can run the command:

python3 optimizedSD/optimized_txt2img.py --prompt "hello world" --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

Stable Diffusion worked on my visual card with 8 GB RAM (alas, I did not behave well enough to get NVIDIA A100 for Christmas, so 8 GB GPU is the maximum I have;).

Running Stable Diffusion without GPU

If the GPU does not have enough RAM or is not CUDA-compatible, running the code on a CPU will be 20x slower but better than nothing. This unauthorized CPU-only branch from GitHub is easiest to obtain. We may easily edit the source code to use the latest version. It's strange that a pull request for that was made six months ago and still hasn't been approved, as the changes are simple. Readers can finish in 5 minutes:

• Replace if attr.device!= torch.device(cuda) with if attr.device!= torch.device(cuda) and torch.cuda.is available at line 20 of ldm/models/diffusion/ddim.py ().

• Replace if attr.device!= torch.device(cuda) with if attr.device!= torch.device(cuda) and torch.cuda.is available in line 20 of ldm/models/diffusion/plms.py ().

• Replace device=cuda in lines 38, 55, 83, and 142 of ldm/modules/encoders/modules.py with device=cuda if torch.cuda.is available(), otherwise cpu.

• Replace model.cuda() in scripts/txt2img.py line 28 and scripts/img2img.py line 43 with if torch.cuda.is available(): model.cuda ().

Run the script again.

Testing

Test the model. Text-to-image is the first choice. Test the command line example again:

python3 scripts/txt2img.py --prompt "hello world" --plms --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

The slow generation takes 10 seconds on a GPU and 10 minutes on a CPU. Final image:

Hello world is dull and abstract. Try a brush-wielding hamster. Why? Because we can, and it's not as insane as Napoleon's cat. Another image:

Generating an image from a text prompt and another image is interesting. I made this picture in two minutes using the image editor (sorry, drawing wasn't my strong suit):

I can create an image from this drawing:

python3 scripts/img2img.py --prompt "A bird is sitting on a tree branch" --ckpt sd-v1-4.ckpt --init-img bird.png --strength 0.8

It was far better than my initial drawing:

I hope readers understand and experiment.

Stable Diffusion UI

Developers love the command line, but regular users may struggle. Stable Diffusion UI projects simplify image generation and installation. Simple usage:

• Unpack the ZIP after downloading it from https://github.com/cmdr2/stable-diffusion-ui/releases. Linux and Windows are compatible with Stable Diffusion UI (sorry for Mac users, but those machines are not well-suitable for heavy machine learning tasks anyway;).

• Start the script.

Done. The web browser UI makes configuring various Stable Diffusion features (upscaling, filtering, etc.) easy:

V2.1 of Stable Diffusion

I noticed the notification about releasing version 2.1 while writing this essay, and it was intriguing to test it. First, compare version 2 to version 1:

• alternative text encoding. The Contrastive LanguageImage Pre-training (CLIP) deep learning model, which was trained on a significant number of text-image pairs, is used in Stable Diffusion 1. The open-source CLIP implementation used in Stable Diffusion 2 is called OpenCLIP. It is difficult to determine whether there have been any technical advancements or if legal concerns were the main focus. However, because the training datasets for the two text encoders were different, the output results from V1 and V2 will differ for the identical text prompts.

• a new depth model that may be used to the output of image-to-image generation.

• a revolutionary upscaling technique that can quadruple the resolution of an image.

• Generally higher resolution Stable Diffusion 2 has the ability to produce both 512x512 and 768x768 pictures.

The Hugging Face website offers a free online demo of Stable Diffusion 2.1 for code testing. The process is the same as for version 1.4. Download a fresh version and activate the environment:

conda deactivate  
conda env remove -n ldm  # Use this if version 1 was previously installed
git clone https://github.com/Stability-AI/stablediffusion
cd stablediffusion
conda env create -f environment.yaml
conda activate ldm

Hugging Face offers a new weights ckpt file.

The Out of memory error prevented me from running this version on my 8 GB GPU. Version 2.1 fails on CPUs with the slow conv2d cpu not implemented for Half error (according to this GitHub issue, the CPU support for this algorithm and data type will not be added). The model can be modified from half to full precision (float16 instead of float32), however it doesn't make sense since v1 runs up to 10 minutes on the CPU and v2.1 should be much slower. The online demo results are visible. The same hamster painting with a brush prompt yielded this result:

It looks different from v1, but it functions and has a higher resolution.

The superresolution.py script can run the 4x Stable Diffusion upscaler locally (the x4-upscaler-ema.ckpt weights file should be in the same folder):

python3 scripts/gradio/superresolution.py configs/stable-diffusion/x4-upscaling.yaml x4-upscaler-ema.ckpt

This code allows the web browser UI to select the image to upscale:

The copy-paste strategy may explain why the upscaler needs a text prompt (and the Hugging Face code snippet does not have any text input as well). I got a GPU out of memory error again, although CUDA can be disabled like v1. However, processing an image for more than two hours is unlikely:

Stable Diffusion Limitations

When we use the model, it's fun to see what it can and can't do. Generative models produce abstract visuals but not photorealistic ones. This fundamentally limits The generative neural network was trained on text and image pairs, but humans have a lot of background knowledge about the world. The neural network model knows nothing. If someone asks me to draw a Chinese text, I can draw something that looks like Chinese but is actually gibberish because I never learnt it. Generative AI does too! Humans can learn new languages, but the Stable Diffusion AI model includes only language and image decoder brain components. For instance, the Stable Diffusion model will pull NO WAR banner-bearers like this:

V1:

V2.1:

The shot shows text, although the model never learned to read or write. The model's string tokenizer automatically converts letters to lowercase before generating the image, so typing NO WAR banner or no war banner is the same.

I can also ask the model to draw a gorgeous woman:

V1:

V2.1:

The first image is gorgeous but physically incorrect. A second one is better, although it has an Uncanny valley feel. BTW, v2 has a lifehack to add a negative prompt and define what we don't want on the image. Readers might try adding horrible anatomy to the gorgeous woman request.

If we ask for a cartoon attractive woman, the results are nice, but accuracy doesn't matter:

V1:

V2.1:

Another example: I ordered a model to sketch a mouse, which looks beautiful but has too many legs, ears, and fingers:

V1:

V2.1: improved but not perfect.

V1 produces a fun cartoon flying mouse if I want something more abstract:

I tried multiple times with V2.1 but only received this:

The image is OK, but the first version is closer to the request.

Stable Diffusion struggles to draw letters, fingers, etc. However, abstract images yield interesting outcomes. A rural landscape with a modern metropolis in the background turned out well:

V1:

V2.1:

Generative models help make paintings too (at least, abstract ones). I searched Google Image Search for modern art painting to see works by real artists, and this was the first image:

I typed "abstract oil painting of people dancing" and got this:

V1:

V2.1:

It's a different style, but I don't think the AI-generated graphics are worse than the human-drawn ones.

The AI model cannot think like humans. It thinks nothing. A stable diffusion model is a billion-parameter matrix trained on millions of text-image pairs. I input "robot is creating a picture with a pen" to create an image for this post. Humans understand requests immediately. I tried Stable Diffusion multiple times and got this:

This great artwork has a pen, robot, and sketch, however it was not asked. Maybe it was because the tokenizer deleted is and a words from a statement, but I tried other requests such robot painting picture with pen without success. It's harder to prompt a model than a person.

I hope Stable Diffusion's general effects are evident. Despite its limitations, it can produce beautiful photographs in some settings. Readers who want to use Stable Diffusion results should be warned. Source code examination demonstrates that Stable Diffusion images feature a concealed watermark (text StableDiffusionV1 and SDV2) encoded using the invisible-watermark Python package. It's not a secret, because the official Stable Diffusion repository's test watermark.py file contains a decoding snippet. The put watermark line in the txt2img.py source code can be removed if desired. I didn't discover this watermark on photographs made by the online Hugging Face demo. Maybe I did something incorrectly (but maybe they are just not using the txt2img script on their backend at all).

Conclusion

The Stable Diffusion model was fascinating. As I mentioned before, trying something yourself is always better than taking someone else's word, so I encourage readers to do the same (including this article as well;).

Is Generative AI a game-changer? My humble experience tells me:

• I think that place has a lot of potential. For designers and artists, generative AI can be a truly useful and innovative tool. Unfortunately, it can also pose a threat to some of them since if users can enter a text field to obtain a picture or a website logo in a matter of clicks, why would they pay more to a different party? Is it possible right now? unquestionably not yet. Images still have a very poor quality and are erroneous in minute details. And after viewing the image of the stunning woman above, models and fashion photographers may also unwind because it is highly unlikely that AI will replace them in the upcoming years.

• Today, generative AI is still in its infancy. Even 768x768 images are considered to be of a high resolution when using neural networks, which are computationally highly expensive. There isn't an AI model that can generate high-resolution photographs natively without upscaling or other methods, at least not as of the time this article was written, but it will happen eventually.

• It is still a challenge to accurately represent knowledge in neural networks (information like how many legs a cat has or the year Napoleon was born). Consequently, AI models struggle to create photorealistic photos, at least where little details are important (on the other side, when I searched Google for modern art paintings, the results are often even worse;).

• When compared to the carefully chosen images from official web pages or YouTube reviews, the average output quality of a Stable Diffusion generation process is actually less attractive because to its high degree of randomness. When using the same technique on their own, consumers will theoretically only view those images as 1% of the results.

Anyway, it's exciting to witness this area's advancement, especially because the project is open source. Google's Imagen and DALL-E 2 can also produce remarkable findings. It will be interesting to see how they progress.