Home »Forum»Open mic at the blue Iguana »AI-tools/tools for guitarists

AI-tools/tools for guitarists

posted on #1
Member
Posts: 246
Joined: 19.08.13
We all know there is a new industrial revolution coming based on artificial intelligence.

When will it touch the life of musicians?

I'd wish I had a simple listening tool that tells me the bpm I use in its display. This function could be part of my little pcm-recorder. Ok, the AI-circuit has to know the beat I use, e.g. 3/4, 6/8. In 3/4 I need the tempo in quarter-notes and in 6/8 I need the tempo in dotted quarters because we are talking about music not volume-events.

The next device I'd need is a little robot who sits besides me and print my tabs in real time. I expect he can record what I do, hears it through his mics, can replay it like a very talented musicians and can tab it 1000 times faster than I can do - if I can :-)

It would be handy if this little robot could create drum-files like the best Wikiloop-drummer can do, just by listening to a recording or playing live together.

Yet, I don't have this bpm-shower, the tabber-robot and much more tools.

The best tool I use is a Shubb-Capo. Very, very useful! The best device I ever bought. The best device to get creative is a little pcm-recorder. Don't know why but it works in a incredible way because it is so easy to use as an audio-diary. A strong buy for everyone who wants to get creative!

Do you actually use a audio to notation-converter that really got some intelligence/knowledge on its own? Do you know a tool to show the bpm in real-time like a stage-tuner?
Edited by Neronick on 29-09-2016 10:12
Was born in an analog world.
posted on #2
Supporter
Posts: 337
Joined: 25.03.12
These tools will NOT be around for any time beeing. And the reason for that is quite simple. AI research has taken a direction that is mostly oriented towards something that could best be described as glorified statistics. True vectorisation is very sparse

Take Shazam for example. It does what it is supposed to do and it does it quite well. That is identify one specific recording (and that works best on material with strong rhythmic content), but to my knowing it is not able to identify a melody or a chord progression.

There are audio to midi converters that work. But these work only in a very well defined domain. Until recently conversion had to be monophonic forcing guitarists to mount hexaphonic pickups or buy special midi guitars. Now there is a software called [url=http://www.jamorigin.com/products/midi-guitar/] Midi Guitar[/url] that is surprisingly useful. But it still has to be used in a very well defined domain. Any effects or non guitar sounds render it useless. All program that i have tried that claim to be able to do a general audio vectorisation are complete and utter crap. Probably because they take the wrong approach to the problem leading to bad and bloated algorithms and having high demand on hardware.

The Musical OCR programs i have tried do not work either. The situation is about the as for text OCR 25 years ago. That is in most cases you get a result much quicker by entering the information by hand then by scanning, OCR'ing and correcting the errors. And as i see it, OCR is a much simpler task than audio vectorisation.

When it comes to BPM identification i have spent some considerable amount of time searching for a useful envelope detection algorithm that is capable to separate the attack and decay regions of a musical recording without success. Such an algorithm is crucial to BPM classification IMO. All algorithms i found were written from a view of a schoolbook physics model which is oversimplified and bears little or no understanding of what makes up actual musical content.

AI is very far from what salespeople and hollywood wants us to believe. The only areas where is produces semi decent results are fields that are oriented towards economics and marketing and where there is big data available
Pure fingerstyle
posted on #3
Member
Posts: 246
Joined: 19.08.13
Thank you for your detailed view, Nilton. I have no personal experience what is the actual edge of science about AI. All I know it does not happen in music but they candrive big vans without a driver. They can survive the growing of plants with autonom units which decide with "intelligence...

The actual book I enjoy has the title"Die Welt in 100 Jahren", written 1910! In no way by fools! A short summary of all their thoughts could be: OUR thoughts will do the same mistakes when it comes to future as these scientifics had done (sorry for wrong english grammar by the way)!

Our brains are clued to our actual world. ALL predictions will fail. Maybe all GPS-satellites will be destroyed in one second by the sun? Cars won't fly but airplanes will stop.

But I bet that the field of music could be one of the effectives fields to improve the "first" generation to come of what we call artificial intelligent.

There is of course no intelligence implemented in a scoring software, even if you use a midi keyboard for input.

Ok, the first thing to work on will be to get the right thresholdlevel for my strummings, then I generate a midi-event and will get the slash notation for my eyes after some work.

Thanks for helping! Sorry for my mobile one-finger-touch-display with lots of mistakes... :)
Edited by Neronick on 30-09-2016 08:50
Was born in an analog world.
posted on #4
Supporter
Posts: 337
Joined: 25.03.12
You are absolutely right about our brains being clued to the outside world. But IMO this is done by our brains actually simulating the world and our interactions are with this simulated model, not the real world per se. Our sensory input is foremost a means of updating this model, not something that is acted directly upon except in emergency situations.

That means that our view of the world is totally biased by the structure of our internal model. That's why our predictions fail amongst others. (The best example would the comment of a Swedish politician who stated that "the internet is just a passing trend"!!. So do not take my predictions too seriously.

There are a number of mechanisms that could break a trend. Mostly the enthusiasm of individuals. Today's global communities make a perfect hotbed for that. And together with crowdfunding this can spin of some very exciting development
Pure fingerstyle
posted on #5
Member
Posts: 246
Joined: 19.08.13
Ok, I am off-topic now but I wouldn't share your enthusiasm about the internet-hype to build a better world. In contrary this is the best and latest example how to spoil and abuse a new idea to old greed.

The rich are getting rich and the poor are kept poor. Maybe you know my song with these lyrics. :)

The middle class fades away. Have a look at Frankfurt/Main!

The human race can build rockets, cars, tunnels and bridges. They couldn't find a solution for the social question in the last 10000 years and it will not be solved in the next 2000 years to come.

The internet is a tool to keep the social question unsolved, too.

It's a new idea of an old brain. No reason for any hope of a better world. Let's face the facts and do the best with it. :)
Was born in an analog world.
posted on #6
Supporter
Posts: 337
Joined: 25.03.12
It has not much to do with the internet itself but with the ability to share ideas and give feedback. And you are right that it mostly is gainful for the already rich.

But there are new mechanisms emerging enabling people to develop their ideas outside the established structures. The whole maker movement is a good example. Sadly enough this has attracted so much attraction from the capital/establishment so that it tries to capitalize on this movement and thus influencing it very negatively. But nonetheless the movement is there and very interesting ideas keep emerging.

All this will lead to new and useful products eventually, and many of them developed outside the current establishment of universities and product development. But this development is more dependent on enthusiasm and the will to share than on capital resources and therefore much harder to predict. Overall development will also be slower than in the commercial world except for hotspots, but quality will often be higher.

I think one of the best examples of this is the Linux operating system.
Pure fingerstyle
posted on #7
Supporter
Posts: 61
Joined: 25.08.14
Not very accurate and not in real time is Riffstation, I use it sometimes for notating very fast and atonal licks cause it has a slowdown function.
In most DAWS, but also not in real time, you can render audio to midi and from midi to standard notation is an easy one,
Melodyne does a better job but all have troubles with polyphony.
posted on #8
Supporter
Posts: 337
Joined: 25.03.12
This very much confirms my statemens before.
If you need a tool with good slowdown and transposition functionality i can recommend [url=https://www.seventhstring.com/xscribe/overview.html] Transcribe [/url]

Another good example of development off mainstream
Pure fingerstyle
posted on #9
Supporter
Posts: 61
Joined: 25.08.14
didn't know bout that one Nilton,sure gonna try that one out,thanks!
wikiloops online jamsessions are brought to you with friendly support by:
anonymous

"wikiloops has helped so much in my quest to learn. such a great site. and thanks to all the musicians out there that have uploaded some awesome music."

wikiloops.com uses Cookies to provide you the best possible browsing experience.
Read more in our data privacy policy.