Where "real" intelligence has failed, perhaps "artificial" intelligence will work!
I'm fairly certain I have railed on this subject in the past, but to no avail, so I will try again. It would make logical sense to me to have the Style Picker actually select a style that was suited to the intended song. But alas, it doesn't seem to "make the connection". (Using Version 2024, 1113) 64 bit.
Here's the process I'm using: 1) Open the MuliPicker Library using the icon 2) Select the Styles tab 3) In the box "Type in a familiar song title ...", I type in "Somewhere over" 4) Up pops 3 suggestions, I pick the 3rd one by "Izzy" and click "Enter" 5) The Picker selects 451 choices from my list of 8,112 6) Since my "Type" is set to "All", it of course lists the RealStyles first, then... 7) MIDI styles are listed next
So here's the problem as I see it. I assume that at some point, a "real" person had actually listened to the original song, and within one measure, should have figured out that the style required, had to be something to do with a ukulele.
In looking at the list the Picker pulled up, no uke styles in the first 10 RStyles, and none in the first 10 MIDI styles, in fact, I don't see a uke style in the list at all, despite knowing that there ARE some in my Styles folder.
Sooo, I am hoping that in a new, updated version of the StylePicker, there will be some form of AI applied to figure out you need a uke style for a song with a uke in it. Not rocket science.
Think it's an isolated case? Try typing in a Johnny Cash song title or an Everly Brothers song title and see if any styles come up with those in the name. Just wondering.
I've pretty much given up on this feature! It has never yielded useful results for me. And it is just annoying to see so many useless results offered up as if they were a match.
I've pretty much given up on this feature! It has never yielded useful results for me. And it is just annoying to see so many useless results offered up as if they were a match.
That's mainly been my experience also.
I like the suggestion that a more AI approach might find better matches.
BIAB & RB2024 Win.(Audiophile), Sonar Platinum, Cakewalk by Bandlab, Izotope Prod.Bundle, Roland RD-1000, Synthogy Ivory, Kontakt, Focusrite 18i20, KetronSD2, NS40M Monitors, Pioneer Active Monitors, AKG K271 Studio H'phones
https://soundcloud.com/user-646279677 BiaB 2024 Windows For me there’s no better place in the band than to have one leg in the harmony world and the other in the percussive. Thank you Paul Tutmarc and Leo Fender.
haveing worked in software development in the past. this is deja vu for me. and its perfectly understandeable that people want features in tech like AI etc etc in the hope of makeing their lives easier.
but i guess i'm a bit of a AI sceptic. lets remember underlying AI is just programming code. thats all it is. ie a creation by people. yes the coders can do some pretty fancy things these days like intelligent searches but it will only take one so far.
for me ive tried the search features and like the rest of yoiu not been particularly happy with the results. but i would argue that pg coders can only go so far. so...on a serious song rather than a fool about i always have to go through a ton of styles and pg content often to find what i hear in my head. for example in my pretty girl song in my sig i had this sound in my head for a kind of little tremelo-ish 50's-ish strum at a couple of time points in the song. it took hours of work to find that...but once i did...it was pure joy to find it. songs are just work.
bottom line with great respect to everyone here i feel AI will only take one so far. just my opinion as its all code. maybe i'll get egg on my face in the coming years...lol.
happiness to all.
om
Last edited by justanoldmuso; 08/29/2403:58 AM.
my songs....mixed for good earbuds...(fyi..my vocs on all songs..) https://soundcloud.com/alfsongs (90 songs created useing bb/rb.)
I would hope the "sorting on multiple columns" is already in a separate thread. I'm onboard for a +1 on that. However, I don't believe any amount of sorting would find a uke style in my example. The underlying problem as I see it is, that the person who assigned the attributes to the "Izzy" example, didn't bother to, or had no way to, take into account and enter an attribute for, a uke sound.
Now I confess to know very little about what A.I. can do. But seeing and using software like "Song Master" that can separate vocals, drums, bass, etc., it seems to me that turning my sample song over to AI it would have figured out there was a uke there, then a quick search through the BIAB Styles list would have picked up a few, and posted them at the top of the search list.
The underlying problem as I see it is, that the person who assigned the attributes to the "Izzy" example, didn't bother to, or had no way to, take into account and enter an attribute for, a uke sound.
The song title descriptions for "Somewhere over the Rainbow" seem accurate (see attachment). Although I am wondering if it is actually a swing tune - it's a bit in between.
The style filters are set based on feel, tempo, genre of the song, and perhaps it could place more emphasis on keywords - e.g. "ukulele" as you suggest. There are almost certainly some options we could add to improve the feature using the existing information in the database that has manually been created by people listening to the tunes.
What are the top styles that YOU would have chosen, had you chosen the best one for the IZ version of Somewhere...? It would probably be best just to start with a blank style and look for a Ukulele RT that might work. I mean, the best one is probably FolkSwing 140... would you say?
Hi Lloyd, I followed your 7 item workflow and made a single change and my StylePicker list of 11,165 filtered down to 11 and also found styles with RealTrack ukulele. The difference was I used the Filter String box to add ukulele to my search. Using the filter String search box, I also found styles for Johnny Cash and the Everly Brothers.
I'm also not in favor of AI style searches. I would hate to use a style and then get a cease and desist letter of threat of a lawsuit from Taylor Swift's legal team because the style AI found had two measures of a Taylor Swift song I never heard. I don't think AI is ready for primetime yet.
Maybe, but ignoring it may be our own peril. I'm certainly not a fan mainly because I am concerned that it will just be used by many to dumb our own intelligence, but it's here to stay, and will continue to provide influence.
Quote
I would hate to use a style and then get a cease and desist letter of threat of a lawsuit from Taylor Swift's legal team because the style AI found had two measures of a Taylor Swift song I never heard.
That cannot happen with the current StylePicker search functions?
BIAB & RB2024 Win.(Audiophile), Sonar Platinum, Cakewalk by Bandlab, Izotope Prod.Bundle, Roland RD-1000, Synthogy Ivory, Kontakt, Focusrite 18i20, KetronSD2, NS40M Monitors, Pioneer Active Monitors, AKG K271 Studio H'phones
Rustyspoon: my thought is that AI should be employed when the attributes are assigned to a particular song, and that would be well before it was published to BIAB. It would be done at the same time that the attributes: Lite Pop, 4/4, Ev16 etc. were assigned to the song, so that when the search was done, it would list the Uke styles at or near the top. You wouldn't need the internet at that point.
Andrew: absolutely the style filters are correct. But no matter how accurate they are, they won't pick up the most appropriate style. Perhaps the Style db (or whatever is being used) just needs more fields, or columns. Sure you could filter further as you (and others here) have suggested, but that doesn't solve the underlying problem. I looked up my BIAB file for this song. The style shows "_UKES140.STY".
Charlie: (see my response to Rusty above). The AI would already have happened, done at the time the attributes were assigned, NOT at the time you went to pick a style.
All I'm trying to point out here is, there's obviously a REAL person involved when a song is added to the BIAB db. That person has to listen to the song, by the ascribed artist, and make a determination as to what attributes need to assigned, so that the Style Picker has something to work with. Why not use AI at that point and have it figure MORE about the song to make a better determination as to what Style would work.
Andrew: absolutely the style filters are correct. But no matter how accurate they are, they won't pick up the most appropriate style. Perhaps the Style db (or whatever is being used) just needs more fields, or columns. Sure you could filter further as you (and others here) have suggested, but that doesn't solve the underlying problem. I looked up my BIAB file for this song. The style shows "_UKES140.STY".
All I'm trying to point out here is, there's obviously a REAL person involved when a song is added to the BIAB db. That person has to listen to the song, by the ascribed artist, and make a determination as to what attributes need to assigned, so that the Style Picker has something to work with. Why not use AI at that point and have it figure MORE about the song to make a better determination as to what Style would work.
LLOYD S
+1 This is the key issue! The current filters does not catch the characteristics/dimensions of the song at hand that gives it it’s specific feel/groove (i.e. what you hear in your head and what you would recognize as very similar in the underlaying groove/feel/characteristics).
The current filter parameters only produces a list of styles with the same filter parameters but without any ressemblance to the feel/groove of the original song.
In order for this to be of any use there need to be another "dimension - attributes/parameters" that catches the essence of the feel/groove of the song at hand (I don’t have any clue what that would be, but it is absolutely needed on order for this function to be of any use at all). I think one dimension that is missing is the ability to search from a rhythm perspective building upp the underlaying groove, characters of drums, bass and rhythm played by the rhythm instrument.
I would guess that the person listening to the song only listening to find and set the basic filter parameters nothing else because there is probably nothing more to filter on in the system (may be to put in a text string as suggested above). And it is probably far to much work to go through the resulting list and mark up any good matches (the same work as the users are expected to do).
It is very useful to have the filter parameters set (but I could also set them myself). What I really need the system helping me with is to take it from there.
It is obvious that there is a need to characterize a song in more dimensions in order for this to be of any use. It is quite easy to know when listening to a style that ressembles well with the song at hand that it is a good match but I don’t have a clue what makes it up, Of cause it is the basic filter parameters in the bottom but there are so much more needed in order to hear it as a good match.
I don’t have a clue what it could be but maybe other users could suggest what is needed in order for a good match. I don’t think it is the number of instruments or type of instruments as such one can often hear that there is a good match in the underlaying groove/feel. (How much can be removed from a style or a song and still have a good match).
Much of the potential power of BIAB and its huge music library falls flat if the users can't easily find what they are looking for.
A good point is being made here. Someone had to assign key descriptive words to a style.
During beta testing, if I see some description that is clearly wrong, I report it. Usually that is an indexing error, rather than a judgment call. But most of the time I don't do this, especially after release. For example, my definition of jazz centers on modern, Brazilian, Latin, big band etc. but someone else may think Dixieland or Smooth or Funk or Blues etc. when jazz is mentioned. Both approaches are correct to call 'jazz' but having both will give lots of suggested styles that may not be helpful to an individual user.
BIAB 2024 Win Audiophile. Software: Studio One 6.5 Pro, Swam horns, Acoustica-7, Notion 6; Win 11 Home. Hardware: Intel i9, 32 Gb; Roland Integra-7, Presonus Studio 192, Presonus Faderport 8, Royer 121, Adam Sub8 & Neumann 120 monitors
I agree with Lloyd and shlind; there is a deep need to be able to find, fine-tune and create styles that catch the characteristics of the song one has in one’s head.
Shlind talks about the need for another “dimension” that “catches the feel/groove of the song”. This is certainly one way to think about the solution. But I too have no idea what that other dimension could be. Does anyone?
As an engineer I have to look at this from an “engineering perspective” and decompose the problem into two parts, a Part A and a Part B.
Part A As I see it, Part A is to acquire a deep and working understanding of what music is. For example, what is it that makes any artist sound like themselves? Is it the drums? No. There are lots of artists that use the same patterns. Guitar? No, for the same reason. Synths? No. Bass? No. Percussion? No. Vocals? No. Effects? No. Unique Skill? No. For sure, all of these are important but more is needed to truly replicate any given artists song.
The detailed “recipe” that the artist in question uses appears to be a fundamental question that must be answered at a deep and complete level for any algorithm or AI to successfully replicate the song. We use terms like “feel” and “groove” but what do they really mean? These terms need to be defined sonically and perhaps mathematically so that the programmers can program.
The good news is that skillful, discerning musicians instinctively know the recipe even if they can’t articulate it without their instruments. Case in point is the Foxes and Fossils cover of Sail on Sailor. Not only were they able to replicate the original to a very high level (not to be confused with duplicate), but I happen to like their rendition better than the Beach Boys original! Another example of this for me is Killing Me Softly by Roberta Flak. Most likely everyone has their own examples.
Much like a world-class chef uses the same elements as us lesser cooks (meat, vegetables, grains, water, spices, temperature, time, etc.) it’s the recipe (or process) that makes the difference. Given that the human tongue can detect only 4 tastes; sweet, salty, sour, umami and bitter, how can it be that in the course of just 1 week we can detect dozens of different flavors and over a lifetime, hundreds if not thousands? The answer is a recipe of combinations of tastes. So too with music.
More good news is that much of the heavy lifting has already been done in this arena and there are vast resources available. One resource is Librosa , a Python library. This is not at all to say that this is an easy problem to solve. Quite the contrary, but eventually some team will eventually solve it.
Part B The Part B of the solution (imho) is taking the progress and knowledge from Part A and exploit the wonderful RTs and RDs to create a general solution using software and perhaps machine-learning tools. It seems to me that the HIL (Human-in the-Loop) is the weak point. If humans are painstakingly categorizing popular songs into style buckets it’s likely this is not sustainable because 1. There are far too many songs to categorize; over 100,000 songs are uploaded per day worldwide 2. New genres will continue to be created by the movers and shakers in the music industry; genres that could not have been recorded when the RTs and RDs were recorded because they didn’t exist at that time. 3. Users may want to replicate unpopular songs that were never categorized in the first place.
For grins, I asked the following question to my AI assistant. Given a static and limited set of drum loops, guitar recordings, keyboard riffs, and other recordings, is it known by musicologists exactly how to accurately replicate a given song using these recordings?
In addition to the complex task of considering the arrangement, structure, timing, tempo, harmony, melody, instrumentation, timber, mixing and production, this was the summary answer. While musicologists and producers can get very close to replicating a song using these techniques, achieving an exact match involves a deep understanding of the original recording’s nuances and the creative decisions made during its production.
This answer tells me that software is the only way to produce a general and complete solution.
https://soundcloud.com/user-646279677 BiaB 2024 Windows For me there’s no better place in the band than to have one leg in the harmony world and the other in the percussive. Thank you Paul Tutmarc and Leo Fender.
FWIW - I believe that PGM should look at EZKeys2 Bandmate. For those not familiar with Bandmate you would being in a song or chord progression into it and it would select various MIDI loops that would fit. It also gives a percentage of what MIDI loops fit the best. Although it is not perfect it is better than BiaB's process.
Wouldn't it be great if BiaB and RB had a similar function? For example bring in a WAV or MP3 into BiaB or RB and have it list various styles that match and/or comes very close to the WAV or MP3. That would be AI at its best.
YMMV
When I was young I was poor. But after years of hard, honest, and painstaking work I am no longer young.
64 bit Win 10 Pro, the latest BiaB/RB, Roland Octa-Capture audio interface, a ton of software/hardware
... Wouldn't it be great if BiaB and RB had a similar function? For example bring in a WAV or MP3 into BiaB or RB and have it list various styles that match and/or comes very close to the WAV or MP3. That would be AI at its best.
If I recall correctly, Peter once acknowledged this concept, but I'm not sure if it is on the 'todo' list.
BIAB & RB2024 Win.(Audiophile), Sonar Platinum, Cakewalk by Bandlab, Izotope Prod.Bundle, Roland RD-1000, Synthogy Ivory, Kontakt, Focusrite 18i20, KetronSD2, NS40M Monitors, Pioneer Active Monitors, AKG K271 Studio H'phones
Ci siamo dati da fare e abbiamo aggiunto oltre 50 nuove funzionalità e una straordinaria raccolta di nuovi contenuti, tra cui 222 RealTracks, nuovi RealStyles, MIDI SuperTracks, Instrumental Studies, "Songs with Vocals" Artist Performance Sets, Playable RealTracks Set 3, Playable RealDrums Set 2, due nuovi set di "RealDrums Stems", XPro Styles PAK 6, Xtra Styles PAK 17 e altro ancora!
Wir waren fleißig und haben über 50 neue Funktionen und eine erstaunliche Sammlung neuer Inhalte hinzugefügt, darunter 222 RealTracks, neue RealStyles, MIDI SuperTracks, Instrumental Studies, "Songs with Vocals" Artist Performance Sets, abspielbare RealTracks Set 3, abspielbare RealDrums Set 2, zwei neue Sets von "RealDrums Stems", XPro Styles PAK & 7, Xtra Styles PAK 17 & 18, und mehr!
Band-in-a-Box® 2024 apporte plus de 50 fonctions nouvelles ainsi qu'une importante de contenus nouveaux à savoir : 222 RealTracks, des RealStyles nouveaux, des SuperTracks MIDI, des Etudes d'Instruments, des Prestations d'Artistes, des "Morceaux avec Choeurs", un Set 3 de Tracks Jouables, un Set 2 de RealDrums Jouables, deux nouveaux Sets de "RealDrums Stems", des Styles XPro PAK 6 & 7, des Xtra Styles PAK 17 & 18, et bien plus encore!
New! XPro Styles PAK 7 for Band-in-a-Box 2024 for Mac!
We've just released XPro Styles PAK 7 with 100 brand new RealStyles, plus 50 RealTracks and RealDrums that are sure to delight!
With XPro Styles PAK 7 you can expect 25 rock & pop, 25 jazz, and 25 country styles, as well as 25 of this year's wildcard genre: Celtic!
Here's a small sampling of what XPro Styles PAK 7 has to offer: energetic rock jigs, New Orleans funk, lilting jazz waltzes, fast Celtic punk, uptempo train beats, gritty grunge, intense jazz rock, groovy EDM, soulful R&B, soft singer-songwriter pop, country blues rock, and many more!
Special Pricing! Until September 30, 2024, all the XPro Styles PAKs 1 - 7 are on sale for only $29 ea (Reg. $49 ea)! Supercharge your Band-in-a-Box 2024® with XPro Styles PAK 7! Order now!
XPro Styles PAKs require Band-in-a-Box® 2024 or higher and are compatible with ANY package, including the Pro, MegaPAK, UltraPAK, UltraPAK+, and Audiophile Edition.
New! Xtra Styles PAK 18 for Band-in-a-Box 2024 for Mac!
Xtra Styles PAK 18 for Band-in-a-Box version 2024 is here with 200 brand new styles to take for a spin!
Along with 50 new styles each for the rock & pop, jazz, and country genres, we’ve put together a collection of styles using sounds from the SynthMaster plugin!
In this PAK you'll find: dubby reggae grooves, rootsy Americana, LA jazz pop, driving pop rock, mellow electronica, modern jazz fusion, spacey country ballads, Motown shuffles, energetic EDM, and plenty of synth heavy grooves! Xtra Style PAK 18 features these styles and many, many more!
Special Pricing! Until September 30, 2024, all the Xtra Styles PAKs 1 - 18 are on sale for only $29 ea (Reg. $49 ea)! Expand your Band-in-a-Box 2024® library with Xtra Styles PAK 18! Order now!
Learn more and listen to demos of the Xtra Styles PAK 18 here.
Note: The Xtra Styles require the UltraPAK, UltraPAK+, or Audiophile Edition of Band-in-a-Box®. (Xtra Styles PAK 18 requires the 2024 UltraPAK/UltraPAK+/Audiophile Edition. They will not work with the Pro or MegaPAK version because they need the RealTracks from the UltraPAK, UltraPAK+, or Audiophile Edition.
New! Xtra Styles PAK 18 for Band-in-a-Box 2024 for Windows!
Xtra Styles PAK 18 for Band-in-a-Box version 2024 is here with 200 brand new styles to take for a spin!
Along with 50 new styles each for the rock & pop, jazz, and country genres, we’ve put together a collection of styles using sounds from the SynthMaster plugin!
In this PAK you'll find: dubby reggae grooves, rootsy Americana, LA jazz pop, driving pop rock, mellow electronica, modern jazz fusion, spacey country ballads, Motown shuffles, energetic EDM, and plenty of synth heavy grooves! Xtra Style PAK 18 features these styles and many, many more!
Special Pricing! Until September 30, 2024, all the Xtra Styles PAKs 1 - 18 are on sale for only $29 ea (Reg. $49 ea)! Expand your Band-in-a-Box 2024® library with Xtra Styles PAK 18! Order now!
Learn more and listen to demos of the Xtra Styles PAK 18 here.
Note: The Xtra Styles require the UltraPAK, UltraPAK+, or Audiophile Edition of Band-in-a-Box®. (Xtra Styles PAK 18 requires the 2024 UltraPAK/UltraPAK+/Audiophile Edition. They will not work with the Pro or MegaPAK version because they need the RealTracks from the UltraPAK, UltraPAK+, or Audiophile Edition.
New! XPro Styles PAK 7 for Band-in-a-Box 2024 for Windows!
We've just released XPro Styles PAK 7 with 100 brand new RealStyles, plus 50 RealTracks and RealDrums that are sure to delight!
With XPro Styles PAK 7 you can expect 25 rock & pop, 25 jazz, and 25 country styles, as well as 25 of this year's wildcard genre: Celtic!
Here's a small sampling of what XPro Styles PAK 7 has to offer: energetic rock jigs, New Orleans funk, lilting jazz waltzes, fast Celtic punk, uptempo train beats, gritty grunge, intense jazz rock, groovy EDM, soulful R&B, soft singer-songwriter pop, country blues rock, and many more!
Special Pricing! Until September 30, 2024, all the XPro Styles PAKs 1 - 7 are on sale for only $29 ea (Reg. $49 ea)! Supercharge your Band-in-a-Box 2024® with XPro Styles PAK 7! Order now!
XPro Styles PAKs require Band-in-a-Box® 2024 or higher and are compatible with ANY package, including the Pro, MegaPAK, UltraPAK, UltraPAK+, and Audiophile Edition.
One of our representatives will be happy to help you over the phone. Our hours of operation are from
6:00AM to 6:00PM PST (GMT -8) Monday thru Friday, and 8:00AM to 4:00PM PST Saturday. We are closed Sunday. You can also send us your questions via email.
One of our representatives will be happy to help you on our Live Chat or by email. Our hours of operation are from
6:00AM to 6:00PM PST (GMT -8) Monday thru Friday; 8:00AM to 4:00PM PST (GMT -8) Saturday; Closed Sunday.
Dansk
Deutsch - Windows
Español
Français - Windows
Italiano - Windows
Nederlands
Polski
Português
Svenskt
русский
Japanese
Simplified Chinese
Traditional Chinese
Korean