Recording a Demo Video and Stopping the Google Campaign
After releasing AlbumCoverAI to the public, collecting the first feedback, and improving the UX, I realized that there are some parts that need explaining.
You can only take out so much complexity with UX. In the end, the user needs to understand how the AI works.
There is a prompt; the prompt has many small "criteria" to it, and the AI generates an image that best meets those criteria.
"Lady with pink dress", "cars in the background", and "rainy" are criteria that are obvious to the user. The AI will paint a lady, cars, and rain.
But there are criteria that are implicit to the user but not to the AI: "hyperrealism", "as featured on dribbble", "amazing artwork".
It would seem that the AI generates great images by default, but you must tell it to. You need to provide specific criteria for that. "As featured on dribbble" apparently makes the AI understand that it needs to be a better designer.
Long story short, I couldn't explain this (and other complexities) with UX. I needed a demo video, so users can understand better how the app works if they are stuck. Plus, it provides a glimpse of the app itself - there is no tryout mode, so people can't see the software at work. I could also use a snippet of the video to auto-play on the landing page without audio.
So I recorded that!
...Which was an absolute pain. My Macbook Pro is from 2019 and has an i9 chip ($++), which is one of the worst MacBooks to buy. The i9 chip gets too hot for the size of the laptop, so it overheats and crashes if you even attempt to edit a video.
It took a few screams and some almost-thrown mugs, but after I decided to record snippets no longer than 1-2 minutes, I could stitch something together. My voice struggled with fluency and projecting a sense of excitement, but I'm happy with the result, and I'll get better with practice. You can see the finished video on YouTube.
Next, I needed testimonials from real artists so users could trust the app. This was easy, I just asked my musician friends I interviewed earlier for a quick phrase. Check!
With a longer demo video on YouTube, a shorter showcase video on the landing page, some testimonials, and a few rounds of customer feedback, I was ready to promote the app. I still think it would be better to have a tryout mode that lets users get a feel for the kinds of images the AI generates, but that is too much development effort right now. First, I want to get a few more paying customers.
As mentioned yesterday, I had a Google Ad campaign running, but it yielded little results.
I got about 50,000 impressions and 3,000 page clicks, but only one person bought. This reflected my results from the validation campaign - many page views, no waitlist signups - so I decided to throw the campaign out of the window.
I had a painful learning: I set an end date for the Google campaign so I wouldn't overpay. But I ended up doing just that: A $250 bill arrived yesterday. Painful. A $30 gain on a $250 investment isn't exactly the return I hoped for.
This happened because "setting an end date" made me think the campaign was dealt with. It has an end date; why bother to check it every day? Yes, but the end date still made it run for two weeks.
I have now learned that paid ads should yield results fairly quickly. Otherwise, the ad itself doesn't work. So next time, I will check on the ad daily and pull the plug if it doesn't give me any leads for two days straight.
The meta campaign performed much better during validation, so that's what I'm going to try next. Let's check in tomorrow.