Show HN: We built a voice to music AI app in 54 hours at a hackathon
blog.beatmaticapp.com
blog.beatmaticapp.com
If it continues to works as advertised as it evolves, I think you could quite easily find yourself either in a licensing agreement (or small acquisition) with Apple. Their Garageband app is packed full of "smart instruments" that are in a very similar vein as this. Most of them are quite awesome. My son is obsessed with playing them. They allow even kids to have fun making music and experimenting with sound. Something like this would be a real improvement from what they currently do for "smart" drums.
So, I think there is an initial awesomeness which is that you could have a successful app, which in itself would be a real accomplishment these days. This concept definitely has the potential. However, I really, really hope to see this picked up and incorporated into a larger tool for creating music in a fun way.
Best of luck, guys. Great job.
[1] Might be a bit specific in terms of advice, but not knowing your background/experience building and marketing apps, I thought I'd throw it out there. Coming in with a price tag that's at least $1.99 gives you flexibility with promotions, etc. down the road. Furthermore, something like this definitely provides at least that much value/entertainment.
I've always thought the app has the potential to go viral but it's really nice to hear it from someone else. Thanks also for the advice on not over-engineering and delivering a functioning if minimal product, I'm sure we'd be wise to do that, even though my natural impulse is to add extra functionality! As for the price, I think you're right, in fact Thomas had already suggested it in previous conversations :)
A voice driven smart portastudio would be sublime.
great job and looking forward to see this on the app store!
And yes, Beatmatic can detect any two (and in principle, 3+) "distinct" sounds. We use unsupervised learning techniques (clustering), so your vocal sounds don't need to sound like those used to optimize the software.
As for whether it's AI, the term is a bit subjective. I think it's not unfair to use it as there's a fair amount of statistical techniques and tricks to make it work, and the result is an app doing things you normally wouldn't expect a computer to do.
The 2 minute hack to do what you guys are doing but being able to detect way more that 3 distinct sounds would be to use ableton with the following audio chain..
channel1(audio) -> looper -> vst audio to midi vst -> channel 2 (midi)
set your looper to trigger the on / off recording on channel 2.
Basically a file (currently .wav, but it ought to be .mp3) is uploaded to a the server, which processes it directly in thread and returns the results back in JSON. Really it should be fanned of to a worker pool and update the client with ajax but again, we're not quite there yet.
From there - getting this (beat pattern at the least) into something like Ableton (or some importable format beat format (does that even exist? (Inception!))) would be so awesome.
The biggest thing for me would be for it to be able to handle a pretty complicated rhythm. When i get goin, its a lot faster than your examples (and i'd imagine 2 notes hitting in the same place create a problem as well). I'd imagine it gets pretty complicated pretty quick.
Maybe even a way to record a beat, have it do the conversion, then record over it to make more complicated rhythms.
As you can see i'm excited about your product, and am probably getting way ahead of myself on some of the comments. But either way, keep it up, and let me know when i can try it myself. Thanks!
As far as complicated rhythms go, I'd love to get some audio samples of you beatboxing! One of the things we have planned is a petition for beatbox samples -- we'll write a web app or rich client that makes it as easy and quick as possible. That'll allow us to get some better data than the little corpus I've been building up!
If you want to help out, you can either sign up for updates at http://beatmaticapp.com or email us at dev at beatmaticapp com. It would be much appreciated :)
Anyway, this is 100% Kickstarter material.
I would easily pay $5 just to show off this functionality to friends when we're hanging out (but I agree that it should be kept around the $1.99 price point). The novelty here is huge.
A lot of music apps are intimidating to most people but this is the most intuitive way to take human musical creativity and make a structured, manipulable track out of it. And trust me, most human beings can hum a tune/beat out. It's instinctive so the market for this is huge.
I recommend Kickstarter rather than just shooting for the App Store because it's going to be tough to go viral and get easily discovered on the App Store. Kickstarter, in addition to the money, will get you a solid fan base before you get on the App Store which is really important these days. It's just too easy to get lost among all the other apps, no matter how novel your entry might be.
So get on Kickstarter, raise a decent amount of money to get a very polished app ready, but more importantly, use the time to get your concept viral BEFORE debuting on the App Store. You won't believe how much publicity is generated by the weekly updates that Kickstarter projects provide. Each week you give a little nudge to all the supporters through your product development updates and each week they're reminded to tell friends and family about their backing. The press will of course catch on and you'll get that steam too.
Good luck to you guys. This is definitely at the intersection of liberal arts and technology that Steve Jobs loved touting.
If you guys want to talk more about strategy from here on out, check my user profile to get in touch. I'm definitely supporting this project.
Maybe a way would be to do a 5$ Kickstarter for people to get early access to the app, even before their friends can buy it from the App Store. This could work through the Apple Enterprise Program, but it might not be allowed in the small print...
That could generate some hype, and backers would have something exclusive for a while till it´s available in the app store. It also would build a community that surely gives great feedback and the first few iterations wouln´t need to go through the app store approval process (which would mean faster iterations / feedback).
Either here or on Quora :) http://www.quora.com/Startup-Advice-and-Strategy/We%C2%B4ve-...
Our emails by the way are thomas / martin @ the app domain...
I've been thinking about playing around with that plugin for some time, I just haven't got around to it yet. My understanding is that there is no voice to music feature -- it's more about vocoder and harmonizer effects. It sounds like a genuinely innovative and cool product, though, so I have nothing bad to say about it.
The difference between the two products is that the emphasis of "The Mouth" is on vocal performance, whereas the emphasis in Beatmatic is on composition. The compositional aspect probably doesn't come across that clearly in the videos because, well, all we have implemented is beat detection. But suppose the next step in the app was to voice a bassline, or a melody instrument, or have some cool pre-built synth loops to add to your beat, then we're talking about a music app that can, in theory, offer the ability to write a song to anyone who can sing "Happy Birthday" roughly on key and in time. As a musician who has experienced the joy of writing songs a few times (my band: babylonbaby.de), I'm excited about the idea of sharing that awesome experience with a much larger slice of the consumer market.
But as a professional drummer and tutor, a tool like this could seriously help me, if you decide to expand the feature set beyond a novelty. For example, this could be a way for me to take notes of drum beats that come to me when I'm not at the kit, if it supported tuplets.
You should get to a level to where an average beatboxer can input and this thing spits out something sweet. Game-ify it by having people try to match something onscreen with their mouths or do something a la "draw something" except I input the beat and you guess what song it is or vice versa. Fun!
Congrats, it sounds cool.
I'd buy it just to show my friends and maybe jam a little like you did. So rad. Triple "like". I hope you make a million dollars.
Bad ass application btw. Looking forward to it.