NKS Quality Control
Comments
-
Nope, I am looking in reality and based on how things work now. Not everyone likes/wants drastic change, assistive technologies sure but not as you put it, a whole reimagining.
Anyway, I think we have all made our points, we wont all agree on the same visions, I just think that your suggestions need to be more realistic for the current situation. No issue with adding a simple request for full AI solution, but to pedal it in every thread you see as applicable without any actual technical examples and substance or anything that works in a similar way is a dream, not a request.
Why not dream about an AI plugin that can simply generate all sounds on the planet with synthesis instead of having all this to begin with. Then you can simply tell the machine the sound you want, better yet, hook it up to your neural network and it just does it as you think it, no words needed even.
Better still, just have the AI design and produce the sounds and track, no human needed at all then. Works for video, I can't even tell it was AI generated
0 -
Creepy video... would not drink that beer for fear of DNA mutation.
0 -
That video is sick !!!
This conversation is interesting but indeed tagging seem to be tedious. From a simple user perspective like me, I’d say this whole tagging thing is as it already is a kind of a very good marketing tool simply improving workflow through the use of a NI controller. To me tagging is nice to filter out everything in a quest of something inspiring but having done a bit of tagging myself it lacks controls as I outlined the other day as once created they cannot be removed. I have sometimes random presets that for some reason are read only and cannot be modified (even though I check the file is actually authorized for read/write process) since it makes no sense and sometime is very random and isolated I delete them otherwise I end up with a double library with one containing just the one sound bloating an already pretty filled up list of sample libraries.
Auto tagging would seem a tricky thing to do but certainly after training an AI for some time based on you own way of tagging sounds it should be able to reproduce your own analysis and tag a sound properly based on its transient and frequency/noise analysis and your training criteria. The idea simply put would just be to make the tagging process less tedious. Regarding generating presets from instruments the only tool ever getting close from to achieve something relevant was Preset Magician and it was pretty fast to do so. I was also vouching a lot for being able to drag and drop control pages and add standard keyboard shortcut when editing text field in Maschine making things even more tedious like using tab to jump from a field to another etc as even using cmd+A in a text box doesn’t allow you to select all along with cut copy and paste which seems incredibly wrong from any user perspective.
Though without being too much of a dreamer to go along with the color coding in Maschine you could imagine those being used as well in another for of tagging where each color is assigned to a frequency range and have a colored dot or background based on the dominant frequency which could give the user the ability to jump faster to an area of the list without going through every preview. Another thing that would be great as well (not completely related) since we can hold shift to jump 10 by 10 presets in the list would be to implement the same principle in the routing area (but I guess it is not done for crashing reason with the logo g already consistent) and also most importantly to the plugin selection area as a lot of us have a huge list of plugin and even If I customed my own categories in my user folder I also use a lot browsing by vendor name and have much more than 10 between my purchases, freebees and if like most of us you vote for the KVR challenge every year it adds up quickly 🤓
0 -
I guess AI could possibly learn mimick human tagging by analyzing at least hundreds of thousands of tagged samples. But still human tagging would be better in some cases....
It could be possibly used to find samples/presets that sound similar. If some kind of metrics what similar means exists. Again, it could learn from many users' supplied data. Maybe milions, tens, hundreds of milions of samples.
But still, I see AI way of tagging and human way of tagging will coexist for at least decade, two... It might also be so, that AI will be used internally by companies to do raw tagging that will be later checked for correctness by human just to speed up things and save human resources...
We will see....
1 -
Lots of questions you posed. But most are answered by looking at a current, real-world, in-the-hands-of-millions, example:
Apple Photos Library.
Smart Albums, Keywords, metadata structure, database structure... it's all there to "borrow" from
Your curiosity about "tagless" searching is a pretty fun thing to think about...
Imagine being free of the constraints of "Sound Type" and "Character" as overloaded and jumbled front ends to start a search.
Users could build there own collection of different "Starting Templates" of search criteria.. either with soft keywords or graphic tiles that represent these soft keywords linked to the analyzed NKS2 previews
These simplified and personalized starting templates would lead thru pared-down decision trees.. including some that are not binary, but instead being gradient sliders of characteristics.
You might have a Starting Template of search criteria to build a "Rock Band"
You might have a Starting Template of search criteria to build a "Scary Castle Room"
Users would have fun making these Starting Templates and save time later in building a fresh grouping of presets that work purposely together in a way the user envisions at the time.
The only metadata "burned into" a preset is info about its instrument architecture and available parameters as provided by NKS 2.0. Much like how a jpeg can maintain metadata on Camera Model, Lens, Aperture...
0 -
Apple Photos Library is a pain in the ass - I prefer to create my own organisation not fight with some dumb AI system (which you have to with Apple photo library as it just creates 'albums' without consulting you)
0 -
Apple Photos Library is a good example of why this sort of AI would be a nightmare. I prefer to create my own organisation not fight with some not SMART AI system, which you have to with Apple photo library as it just creates 'albums' and 'collections' without consulting you but they never have what you actually want in them. Even if you add your own you are stuck with it constantly adding its own 'ideas' for albums so you end up with a disorganised mess. So I put any photos I want to organise myself into their own folder and ignore the Apple Photo Library as much as possible.
0 -
Those automatic memories are only one type of grouping..
I can bring together some wonderfully easy groupings with Smart Albums in minutes that would otherwise take hours and still have likely missed some gems.
Who's with who, location, month of the year, pics from a within a small group of years, pics of kids with a favorite pet...all wonderfully easy to do
0 -
Fun to think about creating a "Smart Template" of search criteria...
If you made a "Smart Template" you named "Scary Castle Room".... what first-level prompts would you create to find possible presets for primary elements?
- Ominous low drone
- Tension chord
- Dangerous weather texture
- Creepy crawly melodic
- Shocking Reveal
- Satanic Vocal texture
- Contextual Random
Edit! Just realized this Smart Template needs a prompt for Scary Castle FX too to bring forth the sonic palette!
- Damp Castle Reverb FX
- Dungeon Echo FX
0 -
And then after you click on the first-level prompt you create "Ominous Low Drone", what second-level prompts would you create to actually bring up a group of presets to audition?
- Application Preference: All / Kontakt / Reactor / Massive X
- Instrument Preference: All / Fav List 1 / Fav List 2 / Fav List 3
- Vender/Author Preference: All / Fav List 1 / Fav List 2 / Fav List 3
- Organic <=> Synthetic (gradient slider)
- Perfomance Preference: All / Patterns / Playable
- ?
- ?
- ?
0 -
and, how does one search easily from the keyboard, or if you are blind? You overlook simple things constantly in favor of the big picture. I do not want any of the guff you have listed, just give me simple tags so I can just find a sound. All that sounds like a waste of computer power to me…
0 -
By keyboard, I assume you mean on-device display.
Once you have set up your custom "Smart Templates" for preset searching, they can be driven by simply clicking on your custom tiles/words....
We'll just have to wait and see what's possible on "Mk3" Keyboards.. hmmm...should such a thing ever exist 🙂
And for those with vision issues, voice input to this MUCH smaller custom decision tree vs the monster fustercluck of tagging would be a blessing.
This whole approach much more inviting to a natural language solution.
0 -
Ok... but how does someone that is blind setup a template to begin with?
What about if people dont want to waste time with these so called templates and wants a simple fixed way to find sounds that can be easily memorised and assigned to muscle memory?
Seems it overcomplicates something that is already pretty simple. Knowing how much it would cost to lease that kind of AI tech (working for a company that looked into it for similar reasons but decided against it for numerous reasons, major cost being one of them) I would not believe NI would invest all that for users sake, especially when tagging is already, mostly fine for what it does.
I just think if that is your hope for the next version you will be rather disappointed for the next decade at least.
0 -
I do not expect it at the introduction of "next version" (assuming that a "next version" emerges this Fall). However, I do expect the "next version" will lay down some new technologies that lead to something like this within a year or two.
It is absolutely inevitable...
0 -
Wouldn't hold your breath, but dream if you must.
It is absolutely inevitable...
Death is inevitable, basically the only thing that can be stated with 100% clarity. This I am not sure is on the same scale tho. Anyway, again, point taken, you are passionate if nothing else, just don't expect everyone to agree with the visions you have as it's not something many of us are asking for at all, especially at the cost of performance. I mean we are just talking finding a decent piano sound in a collection of many, and so far I have never felt the need for AI to help me do that.
0
Categories
- All Categories
- 18 Welcome
- 952 Hangout
- 56 NI News
- 471 Tech Talks
- 2.5K Native Access
- 11.1K Komplete
- 1.3K Komplete General
- 3K Komplete Kontrol
- 4K Kontakt
- 846 Reaktor
- 299 Battery 4
- 616 Guitar Rig & FX
- 325 Massive X & Synths
- 768 Other Software & Hardware
- 4.1K Maschine
- 4.8K Traktor
- 4.8K Traktor Software & Hardware
- Check out everything you can do
- Create an account
- See member benefits
- Answer questions
- Ask the community
- See product news
- Connect with creators