#Sample Key Detection
1 messages · Page 1 of 1 (latest)
For long samples, a way to select a region for root note detection (such at a start and length) would be great.
Even a way to select a short region of a sample that will loop so you could use an external tuner.
Edit:
After thinking about this with the compute limitations onboard Move, I think this might work best if it’s made a part of Note and happens when samples are synced via Ableton Cloud?
This is trickier in practice than it might seem, in my experience.
Even the best detection money can buy is limited to major/minor and 12TET: https://mixedinkey.com/learn-more/
All of the other options don't have very good detection rates, like maybe 75% accuracy on average.
One song or sample might not even be in any key/scale, or might have borrowed chords, or might change modes and root notes and so on.
It gets even more difficult when you factor in detuning and glides and bends and so on.
You can always drop a long sample, like a stem, into MiK (which is what I do) to get a good idea of what the key/scale probably is.
For shorter samples, you can use the EQ8, Spectrum, or Tuner utilities - or whatever you prefer for identifying the fundamentals and harmonic overtones and so on - to get a good idea of what you're working with.
Doing all of this on limited compute like what's in the Move would almost certainly not be feasible when it's not even that accurate or flexible on a much more powerful platform.
Yeah, limited functionality would be fine.
I would be happy with being able to select a shirt section of a sample and have it identify the note using an fft, then seeing that as the root for the sample
The more I play with Move, the more I wish something like this was onboard :)
even just the ability to go into a ‘sample tuning’ mode that lets you select a brief loop and plays it while you use an external tuner or something.
Thinking about this some more - if processing power really does completely limit the ability to do this (and I still think there are reasonable ways to make a limited function work, or just loop a short bit to grab a more…)
This could also be a feature that depends on using Note. Capture sample > sync to Note via cloud > key detection occurs on iOS device
Part of the problem is that, no matter how powerful the compute unit is, the results you get may not be very useful.
Yeah, I get what you’re saying. It doesn’t have to be magic, though. It could require the user to select a region and say “this is my root note, what is it?”
Even if it’s just root note detection and not ‘key’ or ‘mode’
What I recommend as part of the workflow with any outboard device that is more limited than a "real computer" is to do this kind of analysis and tagging and renaming and sorting and categorizing and whatever other words might be appropriate to describe "things you should probably do on your computer to make life easier on other platforms" on the "real computer".
Prepare things that require a lot of offline processing, heavy compute, mouse and keyboard, lots of screen real estate, deep menu diving, etc. on the "real computer". Then transform the resources you're working with into "playable" forms. For example, put all of the drums from your favorite sample packs into a "128" drum rack, set up all of the mappings and automation and chains and so on such that you can get quick results out of the drum rack without having to go deeper than the macro knobs and sending MIDI to it, and then save the drum rack in a project or pack.
Then send these "playable devices" to the Push or Move or whatever hardware you've got.
Right, but what I want is to get something good enough to keep sketching without breaking that workflow.
What I mean is, if you have everything already tagged with this info, or you have already configured things such that you already have the info/capabilities you need, then it becomes moot.
I totally get your point about key detection for long samples - but it should be possible to record a sample with the onboard mic, select a short region (a few 100 ms or so), and say “shift the fundamental frequency in this region to the key”
Doesn’t solve all the world’s problems of key detection - but it lets you capture something in the moment and get it close enough to not be totally out of tune with what you’re working on
I understand what you're saying about trying to deduce the fundamental frequency, and thereby the root note, but even this basic operation isn't something that proves very accurate even if you throw a lot of time and money at hardware and software.
that’s fine - it doesn’t have to be terribly accurate. Just better than “way off”
I am interested in what other grooveboxes are capable of in this regard.
I don’t know that any are - it could be a game changer
and the more I think about it, leveraging the ability to offload that to an iOS device could be awesome
Given that the internal compute unit appears to be an off-the-shelf Raspberry Pi v4 with an A72 and 2GB RAM, I am trying my best to resist the temptation to buy a Move and some Raspberry Pi v4 units with higher specs, and see if I can upgrade the stock Move with a better compute unit.
Someone already tried this and didn't get far with the Push 3, but imho they weren't that dedicated or experienced, so I don't think that this sort of thing is impossible or nor merely infeasible.
This is a very interesting idea, not just for the sample note detection use case, but more generally! Like, maybe something that blurs the line between Link and using the Move/Push as a MIDI controller connected via USB cable to a computer. Like a PAN that interconnects your Live-compatible host devices, such as the laptop running Live, the Move, the Push, an iPad running Note, maybe even something like Koala Sampler on an Android tablet... and then offloads tasks from the weaker hardware/software combos onto the "serious" host machines.
This comes to mind:
https://audiogridder.com
Oh yeah, I forgot about audiogridder
I wasn’t thinking of anything happening real-time like that.. just sample, cloud sync, offline processing, sync back results
but that’s a really cool extension of the thought
i def think its waay too technical for the compute on the move but i would love a tuner setting (also for live’s tuner to be more reliable for more timbres but i digress…)
Yeah, I think maybe the best option is something like this: #1294517218763346001 message
thats an interesting option, i tend to use antares autokey app and just play stuff into the microphone and its incredibly accurate enough
Yeah, that’s basically what I’ve been doing so far
Kinda separate from this feature request - but the more I think about this the more I like it hahaha. It would be really cool if low-powered / portable devices had a ‘remote VST host’ kinda like audiogridder which would let you use desktop class plugins (probably for more mixing-type use, since I doubt it would ever keep up for realtime or tracking)