We made $69.95 selling AI-generated music!
Last month, we launched our paywall for our project featuring AI-generated music, Strofe → https://strofe.com.
Strofe makes unique, copyright-free music that can be used in video games, live streams, or even composers who need a base track for their creations:
Even though it's not much yet, me and my friend are both incredibly excited (and a bit in awe) as we weren't sure people would pay for it!
From idea to MVP
We first created an MVP on December last year, and showed it to a few friends and to our network (and investors too, who sometimes have very constructive feedback).
People seemed to like the idea so by April we started to seriously work on it and through FB ads, Google ads we had a bit of traction. The real traction started when we hit a gold mine on very cheap promoted posts on reddit, some of our posts getting 800 + upvotes and almost 40,000 users, so we knew we were on to something.
The flipside of our traffic sources is that bounce rates are high, but still many loyal users have started to like the idea.
Setting up a paywall
We started announcing to our users that we would start charging around September / October, with the chance of everyone keeping their previous song creations for free.
Our model is quite straightforward: you buy credits to purchase and download tracks, for about $4, with cheaper prices if you buy more credits.
We are now experimenting with the social side, giving users free credits for creating and sharing tracks, and so on.
If you wanna try it out, feel free to at https://strofe.com, it's free to make as many tracks as you want. Feedback is very welcome; I'm in charge of all the Front-end development.
Future plans
As the product is fully operational now, we are starting to think about focusing more on the business and marketing part, as well as improving the quality of music, right now it's mostly MIDI based but we will be adding VST's.
We are hoping to be able to generate enough profit so that we don't need to go for the investment route, but of course all roads will need to be considered at some point.
If you have any questions feel free to ask!
Hey YuanHao!
looks amazing! How unique is the generated music?
Thanks arstan! The melody is almost 100% guaranteed to be unique but it can of course sound a bit similar between runs.
For the harmony (IE: accompanying instruments, bazz, drums etc) we keep adding vocabulary, and for every new one added the permutations become exponentially bigger, which we hope will keep creations unique.
Really cool app! Does it do text to music like stable diffusion does with text to image? That would certainly be interesting.
Because text to music is semi-possible with riff diffusion
I was mainly curious if you guys had plans for this, because I'm a part of Evoke, which hosts open source AI models and offers and API for devs making AI apps, so when we end up hosting riffdiffusion, you could potentially implement it into Strofe.
We just have a stable diffusion API for now, but big plans to expand to other AI models.
Also have an active AI discord if you're interested: https://discord.gg/dXJtarPsCm
this looks amazing and worked well for me.
i really like the concept of AI-assisted music, and am interested to see this field expand. have you guys done any work on audio identification using ML? thanks for sharing this