Two Down - Four To Go
So I started the old 'n products in one year' on X for #buildinpublic, six in my case. I started with two ideas: Chat2Print and SupaBuddAi. It's still only January, and I've built two MVP's and launched them, here's what I've learned so far.
This was the first idea I had, and I was pretty excited about it in the beginning because I thought it was a great idea. I still think it's a great idea, but I've learned and realized a lot about the market and the product itself.
It's a print-on-demand service that originally was centred around the browser extension which allows you to export your ChatGPT images directly into your Chat2Print account, then create a mock-up of your design on a product after upscaling it, and then allowing purchase.
Building the MVP
I skipped a landing page, email list and any sort of real idea validation or research. I just went straight to building, the chrome extension, ai product designer, the backend and the frontend. I did this because I was excited about the idea and wanted to build it, I was sure it would be a smash hit with little post launch marketing. Big Mistake.
- Don't build without validating the idea first. I was so excited about the idea that I didn't even think about validating it beyond a couple of keyword searches, not even with an SEO tool just plain ol' Google.
- Don't build without researching the market. I should have posted in subreddits, forums, twitter and other places to see if people would be interested in the product.
- Don't build without a landing page. I should have built a landing page and collected emails to see if people would be interested in the product.
- Don't build without a marketing plan. I should have had a marketing plan in place before I started building the product. I should have had a list of places to post, people to reach out to, and a plan for how I would get the word out about the product.
Nobody is going to buy a monthly subscription to a service that they will use once or twice a year to buy a product. It'll be an impulse buy and/or the side effects of paid ads (which I don't want to do).
Realizing this I streamlined the MVP restricting costly features like image upscaling, storage and access to GPT-4 (DALL-E-3). Okay, great, so I've built all the pieces for the free users although the design output between the two modals is worlds apart.
Stable Diffusion XL Base
A free user can only use the Stable Diffusion XL Base model which is a great model, it's free for me to host as a Cloudflare worker and it gets the job done but it's not DALL-E-3.
Some key differences in comparison to DALL-E-3 are:
- Image quality is far more reliant on user input
- Output dimensions are limited to square (not great for custom phone cases)
- User prompt is not optimized for image generation (not as effective for users with less prompt writing experience)
What does this mean for the free user experience?
It means that free users will have a much harder time creating a design that they like and when they do it will be of lower quality, about 40% of it will be cropped resulting in poor placement on the product. This is a problem.
A paid user will have access to DALL-E-3 which is a much better model especially for beginners. It's a lot more forgiving and the output is much higher quality even with a convoluted minimalistic user prompt, the model rewrites the prompt for safety reasons and to improve prompt quality. It's also optimized for chatbot-like image generation so it's much easier to use.
Some key differences in comparison to Stable Diffusion XL Base are:
- Image quality is far less reliant on user input
- Output dimensions come in square, portrait and landscape (portrait is ideal for phone cases)
- User prompt is optimized for image generation (great for users with less prompt writing experience)
What does this mean for the paid user experience?
Paid users will have a much easier time creating a design that they like and when they do it will be of higher quality, there is no cropping because the upscaled dimensions are big enough that size isn't sacrificed for quality. This is great.
The Technical Stuff
The reason that we lose 40% of the image is because the Stable Diffusion XL Base model outputs a square image, and the product mock-ups are not square so the image is cropped to fit the product mock-up . This is not a problem with DALL-E-3 because it outputs square, portrait and landscape images.
Even after upscaling the image, for instance, a 1024x1024 image upscaled with Real Esrgan will result in a 4096x4096 image, we still lose 40% of the image while trying to achieve the highest DPI (Dots-Per-Inch) possible for the product mock-up, which results in a zoomed in image.
This is not a problem with DALL-E-3 as with portrait images, the dimensions are 1024x1792 which is perfect for the phone case mock-up. After upscaling we have a 4096x7168 image which results in a range of 260 - 300 DPI depending on the phone case dimensions. This is great as anything above 300 DPI has no noticeable difference in quality.
Thinking about this realistically, Chat2Print will flop as it is right now, I may be inexperienced but I'm not stupid. I need to pivot the product to something that will work or I need to scrap it and move on to the next idea.
- I could strip out product purchases making it B2B instead of B2C, an easier sell.
- Pivot the product designer focusing on allowing sellers to export custom designs to their respective sites/stores (Redbubble, Teespring, etc).
- A product marketplace like Etsy, but for AI generated designs. (I don't like this idea, but it's an idea)
- Scrap it and move on to the next idea.
My intention with the challenge was to spend at maximum one month of design and development time on each product and one or two months getting feedback, iterating and marketing. I've spent just over one month in total on Chat2Print and while I could pivot the product, I don't think it's worth it.
I'm going to scrap the idea and move on to the next, I've learned a lot from this experience and I'm excited to apply what I've learned to the next idea. I'll keep track of analytics and see if there's any interest in the product, if there is then I may revisit the idea in the future.
This was the second idea I had, it came to me during week 3 of building Chat2Print. I was anxious about how secure my RLS policies were and I was thinking about how I could make them more secure. Enter SupaBuddAi.
Done It Again
I did it again, I skipped a landing page, email list and any sort of real idea validation or research. I just went straight to building, this was because I never intended to launch it, I just wanted to build it for myself and saw it as something that may be of interest to others.
So What Is It?
The tests are executed in the browser against your Supabase database and the results are displayed in the browser.
- Building the MVP took me about a week as opposed to three for Chat2Print. This was because there was no new APIs to learn, far less testing, optimizing and features to build.
- I took just the landing page and email list and shipped it. No template or GPT-4, just Shadcn, TailwindCSS and a little bit of Typescript.
- I posted on Twitter although that's the only place I posted, I didn't post on Reddit, IndieHackers, HackerNews or any other forums.
- I didn't reach out to anyone, I just posted it and hoped for the best.
An improvement from Chat2Print, but still not great.
No signups, no feedback, no interest. I'm not surprised, I didn't do any marketing, I didn't reach out to anyone, I didn't post anywhere but Twitter.
I'm not going to pivot the product, I'm going to scrap it and move on to the next idea. I'll keep track of analytics and see if there's any interest in the product, if there is then I may revisit the idea in the future.
The Recipe For Success
Researching people who have grown into success through #buildinpublic, I've noticed that they all have one thing in common, their success came after their following. They didn't have a following before they started, they built it through #buildinpublic and then they had success.
Which makes sense, launching without a following is like launching into a void, nobody knows you exist. You need to build a following first, and then launch to that following.
I also noticed that a core component of their success is the speed in which they take an idea from conception to launch. They don't spend months building an MVP, they spend days.
They are able to validate an idea before building anything, it takes them a few days of tweets, posts and emails to validate an idea. I don't have that luxury, yet.
The first two ideas were a flop, they were based on no market research or proven model for success. Post-development SEO research showed very little interest in the products, building content around the products will be difficult and I have no intention of running paid ads.
The next idea which I'll begin building this weekend is based on a proven model for success, it's a proven market, product and business model. It is not an AI or SaaS product, but will be highly SEO optimized from the start.
I still struggle with openly stating my ideas on X, I'm not sure why. I think it's because I'm afraid of someone stealing my idea, but I know that's not a valid reason especially with the ideas and the little followers I've got.
As for #buildinpublic, I need to improve my content strategy. I know I need to post more often, and find my voice with it as I'm not sure what to post about most of the time. So I'll be working on that as well.
If I said anything that is incorrect, of interest, or you have any questions, please reach out to me on Twitter. I would love to hear from you!