I very much appreciate Song Matcher and the other great resources for helping us match BeatBuddy beats to songs. But I’m thinking it would be so nice to use some tech like AI or what we in Pandora, etc., to “listen” to any song we like and suggest the best BB beat for it.
And if we’re feeling fancy, put the right Program change into OnSong so that the beat gets called up automatically when we pull up the chart in OnSong.
But the first thing for sure. Can we do that? Is it already done and I missed it somehow?
Thanks, @persist! I did a similar thing a few months ago with Billy Idol’s Mony Mony. I asked ChatGPT to describe the beat for that song (because I’m a bassist, and sadly I don’t have a drummer’s vocabulary), and among a bunch of other things it said something about “four on the floor.” I had no idea what that was (pretty sure I still don’t), but a search for that came up with the perfect beat for Mony Mony (if you’d like to know, it’s Groove Monkee’s GM_Rock_Essentials_1/Four on the Floor). So that helped, but this is a pretty special case: there’s only one “four on the floor” in my library of BB beats, but there are a ton of rock ones, even rock 8ths ones, which makes ChatGPT’s recommendation for Fortunate Son a little less useful. I haven’t tested that recommendation out for myself, but I work with AI a lot and I have developed a healthy skepticism when it comes to the accuracy of AI-generated output.
Also, I would assume that everybody’s library of beats is a little different, and we need something that will suggest beats from third parties (user contributed, or from commercial vendors) as well as SS content. In other words, we need something that can “know” (by analyzing or looking it up somewhere) about the beats in your BB library of beats regardless of origin, and be able to analyze (or look up somewhere) the characteristics of the song you want to find the right BB beat for, and finally make the comparison and recommendation (then cache the result so it doesn’t have to be computed every time).
I think it might be doable in an old-school way (think Pandora’s Music Genome project), or by new hotness (Gemini, Claude, GPT, or some other more specialized model(s) maybe?).
That would be a feature of Song Matcher I’d pay for, for sure. Anyway, fun idea. Let me know if you ever need a beta tester!
I think AI could actually analyze the song, and via MIDI in BBMO create the beat and grab the drumkit that is closest style wise. I would prefer this over trying to shoehorn a matching beat from the BB library. Just my 2 cents…
I think we all know the next step will be using AI in BBMO to actually write BB beats de novo for whatever song we present it with. Not sure what the copyright issues would be there, though.
I generally avoid using AI (artificial idiotas one of our MIDI sequencers calls it). When I did try to build a MIDI file using ChatGPT, the results were unsatisfactory. It also refused to build one MIDI file, citing copyright restrictions.
I’ve used it for building macros and I can say that over time, the resulting suggestions have been improving.
Currently, you can achieve a similar result using the following tools:
First, download a song and obtain the audio file. You can then use a tool like Vocal Remover to isolate the drums from the track. Once the drums are extracted, you can download the drum-only audio file.
Next, import the drum file into a DAW such as Ableton, which has a feature to convert audio files into MIDI. This allows you to transform the drum audio into a MIDI file.
Afterwards, you can use MIDI Remap to adjust the MIDI mapping from your DAW to match the BeatBuddy mapping. Once this is done, you can create a new song in BBMO (BeatBuddy Manager Online) and import the MIDI drum track.
I’ve tried this process myself. In my experience, it’s not yet perfect, converting drums to MIDI can be inconsistent, especially for more complex patterns or fills.
