Auto-Tagging in Cubase Media Bay (Key, Tempo, Category)

By automatically (sensing) and adding tempo, key and category tags when importing samples into Media Bay and Halion, our workflows would be simplified and we could save a lot of time. Two possible routes to achieve this:

1.) Auto-Write tempo and key info from file name to tags (e.g. tempo, key, category)
Halion 6 already has a feature, whereby you can tell it to take the key from a filename and then map the sample accordingly. I think this option is called “Key Text in Sample Name” and this will set the root key to the key-text that is extracted from the sample file name.

It would be awesome, if this feature could be implemented for auto-tagging WAV and AIFF samples in Media Bay. In which case information would be extracted from the file name and automatically added as tags. Ideally this would include “key” AND “tempo” information, but could also include “category” information taken from the filename (such as drum, percussion, strings, brass aso). It would save everyone a ton of time, as manually tagging files takes ages. And without proper tagging, samples are pretty useless! This feature would be even more powerful, if Steinberg adds:

2.) Auto-Sense tempo, key and genre from file itself (not just filename) and auto-write to tags.
It should also be possible to auto-sense key and tempo from the file itself, even category info could be extracted like this. Audio samples like drums, vox, piano, strings aso. all have very distinctive wave forms - especially one shots). This would definitely be the holy grail of any file browser like media bay. This would be a mega-cool feature for which I would be willing to pay 100 bucks extra anytime. I believe variants of this idea have been requested in the forums before…

In my many years of dealing with large sample libraries, I have seen that One Shots, Hits or Multi-Samples - for example - all have very distinctive wave forms, file sizes and frequency signatures which is certainly true for all drum and percussion hits (lick kick, snare, tom, hi hats, crash cymbals, shakers, claps aso). It should be possible for Steinberg to develop an algorithm, that enables Media Bay to recognize these distinctive file signatures and enables it to automatically add / write the appropriate tags. In essence, this would be an advanced auto-sensing / tagging feature. The Steinberg dev team could also use this feature in upcoming releases of Halion (7), when using the sampler to record audio. Right now, you can record samples which are auto-mapped; Wouldn’t it be cool, if the samples are also automatically tagged AND named?

I believe it should be possible for an algorithm to recognize the most important tags automatically (based on a combinations of file name, file size, wave form, frequency analysis aso). I am sure Steinberg could also license this algorithm to other companies, not just users. Full songs could probably also be categorized automatically this way. I think there is a huge potential for many applications. Many genres could be recognized according to a combination of tempo and instruments: such as EDM (tempo around 125 bpm with kick/bass drum) or Rap (with Vox tempo around 85 bpm) aso.

It would certainly catapult Media Bay - from an already cool information hub - to a whole new level!



Yes PLEASE! This would be a huge improvement!

That sounds really good in theory, like some AI implementation like algonaut or xln xo, but in practice that is hard to be done considering stage where mediabay is now, where in maintenance update(not to mention whole update) it gets broken or some feature that worked suddenly doesn’t work. Apart from general sluggishness which dates way back I witness all kind of odd things with Mediabay through the updates and maintenance updates as well, one of the things stopping me to update on 10.5.20 and 11.
Since my workflow is very depending on mediabay i just hope that they won’t brake and clog it even further, and all stories on the AI wish be only a nice dream.

I don’t want my samples auto-tagged EDM based off BPM.

It would be hard to please everyone, let alone, not piss a lot of people off with a feature like this I think.

Finding Key/Root, Chord Name, etc could definitely be done…

But honestly, there’s so many other simpler things that could be done to improve MediaBay… that I would probably take over AI to which it would be questionable how proficient it would actually be. That’s a longa55 development curve on something to not know how it will pan out.

I’ve tried XLNs and others, and the algos sometimes mix up things like Kicks and Toms, Sound Effects, etc, etc. I need MediaBay to be exactly right, not maybe right… and when I need something more experimental or “fun” or spontaneous, I can use something like XLNs XO or Algonaut

Partially agree here with @LoveGames especially with one that its impossible to please everyone, and that other simpler things that could be improved or done rather than AI.

But, on the other hand its better to have something tagged than nothing, so lets say we are in dream and all is possible, how awesome would be that AI auto tagged samples those tags would be colored yellow instead of white, and have small bottom option to confirm if auto tags are good or partially good (so AI can learn that particular database more efficiently since everyone has different sample content) And those that are tagged manually AI just skips. And also to be able to list some pitch/timbre criteria for ex: high toms separately from low toms or mid toms, or autotag pitch of the vocals(devs like Zplane/Celemony in 99% cases analyses pitch right)etc. And yet some basic stuff like I would have in “character” section more freedom like things such as Low/High/Open/Closed etc so I don’t have to put my manual tags for type of hats/toms etc if they are Open or closed/ Lower or Higher, you got the point.

I had many great ideas nothing with AI that will surely contribute to the Mediabay also I offer those ideas alongside with part of my custom tagged database(on which I am really proud of) all in order to make Mediabay a little more sophisticated than it actually is(and its really powerful) but guys that are charged for Mediabay at Steinberg didn’t show any interest for that.
Now I am not enthusiastic anymore on that I just want that things are worked keeps working.

Luckily tagging system in Mediabay is good enough especially with logical(beside attribute) type of searching. It needs a bit learning curve and some patience for ex I need 4 weeks 4-5 working hours to tag 70-80% here: And after 4 years I am still catch myself that I re-tag or tag something here and there so I am now at 95 of all samples. Today if I had to do it manually again It would took me probably 5 working days to tag those 300k samples.

Whats the benefit?
Enormous! Able to list all egg shakers one-shots or close hats or finger snaps or whatever in a second is a flying rocket. Exploiting all of your sample libs and not only 2-3 while the rest collects dust is awesome and also for songs and creativity as well.

If anything like this was done, it would make a lot more sense for Steinberg to create their own XO or Algonaut that references MediaBay and maybe has - some - data exchange.

And maybe this could also be tied into products like Halion, or be a component of Halion itself - being able to write Halion instrument scripts that access XO type AI database.

I manually tagged over 500.000 samples in Media Bay, which took around 300 hours to do. Thats not creative work. It’s boring and tedious stuff. Recently I’ve been experimenting with various AI driven software to free my time so I can focus on more creative work. And what I have seen in development is very promising and will make Media Bay become outdated very quickly (not to mention that it is very poorly maintained). Of four bug fixes I handed in with Steinberg, only two received a CAN number and were fixed, two were totally ignored. So I stopped handing them in. Feels like a waste of time. Reg Ex is poorly implemented. Won’t go into the details, just look at my previous posts if interested.

Regarding AI: Sorry to break it to you, but AI is a lot more advanced and reliable then you think. The fast pace of innovation is scary. I am currently using Sononym as my number one sample librarian. It’s still in development but is already saving me hundreds of hours in my workflow. When I use it and then switch to Media Bay, I feel like I am leaving a car to sit on a donkey. DAW integration (also for Cubase) is on their roadmap as are many other new features.

Sononym not only has an impressive auto-tagging algorithm (which is continuously being improved), but it makes finding similar sounding samples a breeze. Even if you have absolutely no tags on the sample and a filename like 9309752.wav it will recognize if it’s a one-shoot, snare, kick aso. When you listen to any sample and click on “find similar samples” it actually finds very similar samples (no matter if guitar riffs, melodies, grooves, loops or one shots and and and). Again, this is based on multiple metrics such as file length, frequencies, waveform and more. You don’t need any descriptive filenames, strings or tags. Its all automatic. It’s also non-destructive, because it just maintains it’s own database. To make the bold claim that AI does not work is simply false. Do the research and broaden your horizon. It will also help you save time.

My original post above is actually long outdated. I am using Sonoynm now and I don’t spend hours burning valuable time tagging or doing other useless technical stuff. It’s one of those things that AI is just perfect for. But not only tagging: the way it matches similar samples is really mind-boggling and impressive.

Good luck waiting for Steinberg to discovers AI and implementing it in Media Bay. Anyway, I am getting back into my car as I watch the donkey in my back mirror getting smaller real fast. At the end of the day, everyone gets to decide for themselves how they spend their time. I want to spent it on more creative ways and I will always welcome new technologies, especially if it means I can offload technical maintenance work in favor of creativity. AI is perfect for that.

Yes, Media Bay has Cubase / Nuendo integration going for it. But that is fast changing. And when integration has been overcome by third party apps, then Media Bay will certainly end up on a digital compost heap. Maybe Steinberg will react. Media Bay certainly needs some serious attention in upcoming releases. Bug fixing is not enough.

PS: I wouldn’t describe XO as a sample librarian as it really only has one feature and that is finding one shots. Very narrow usage and the high price can only be a bad joke. But I am sure 2021 will bring some very interesting players (Media Bay will probably not be one of them unfortunately).


You’re making a false claim that I made a false claim. I never made a claim that AI doesn’t work, I said it’s questionable as to how proficient it will be at getting - everything - exactly - right. ie, I don’t want a detuned effected tom drum to be tagged as a kick sample, even if it could potentially be used as a kick sample or “sounds” like a kick sample.

Secondly, AI doesn’t know how I think from day to day. Machine learning is based on what I did yesterday. Sometimes I am tagging samples and material in MediaBay with a certain foresight on how, when, in what, they could be used.

I have specific naming requirements, I’ve tagged just as many as you. Not a single thing is tagged wrong, many have custom descriptions, multiple tags, custom attributes, grouped by family, specific genre, etc, etc. It really hasn’t seemed like a waste of time, because I know my library inside and out, top to bottom , nothing is wasted. Nothing is going unused because of a misstag.

None of these AI programs beat my OCD, at least not yet.

PS, XO - Is a sample librarian, just because it doesn’t have the features you want, doesn’t mean it isn’t.

Its all depending on workflow, I aint detailed explore Sononym as i think when I did several years ago it hasnt get xml read(that u tagged previously in mediabay) and it has its own database(please correct me if I am wrong) which is sensitive if it gets broken while mediabay writes all tags in sample file itself so in case if u move samples to other computer it gets all tags sorted all the way without loss.
Integration is other question here for ex I use this feature all the time which works with 3rd party plugs also:

The paradox is that this trick is not working from 10.5.20 and onward… But thats Steinberg being Steinberg.

I am doubt that Sononym can integrate in similar manner. But please correct me if I am wrong with the mentioned librarian program, and also will be nice if you can share some details about workflow with this, and also does it offers some easy tag migration
from mediabay?

@Grada: Media Bay writes tags into a database and not into the files as far as I know. WAV files cannot accept tags (metadata) as the filesystem does not support it. MP3 and AIFF do support it. There is the option of Broadcast WAV, which can write tags into so called chunks (including iXML), which are basically extensions to the WAV file, but these are not part of a native metadata system to the file itself. If you look into your sample folder, you will see a folder named “.steinberg”. That’s the media bay database. Anyway, just like in Media Bay, you can move your database with your sample folder and both media bay and Sononym should read their respective database with no problem. I use both Sononym and Media Bay concurrently with no problem. I have migrated and changed sample locations multiple times and had no issue.

I did post a questions previously in the forum asking if anyone knew how exactly information is stored in the database or if there is a way to get media bay to convert WAVs to Broadcast WAVs and then write meta directly into the chunk file, but I never received a response. So I guess its a questions that only the dev team of Media Bay can answer.

@LoveGames: I can relate to the OCD with respect to your samples. Its a good feeling to have everything in perfect order. But as Grada rightly put it, it is better to have AI tag your files (with some mishaps) than not to have any tags at all. Also, when AI tags a snare as a tom or vice versa, then I usually see its because the sounds are very very similar (and hence could be used interchangeable in a creative way - if you know what I mean). I also like to follow the Oxford 80/20 rule. Get 80% done quickly, and stop burning time like hay getting the remaining 20% perfect. After all, the goal is to make music and not spend your valuable time maintaining tools. Yes its important, but it shouldn’t become an end in itself. Given the many tools we have now, you can spent hours cataloging everything, but it seems like a never ending project: By the time you finally have everything working, new updates, sample libraries and so on are added and it all starts again. I find that AI can really benefit us here. But I also understand that things are just starting.

I have developed very complex RegEx Expression to search and rename my sample files. Just using Regular Expressions in an advanced way, you can already get a lot done quickly. I use a software called “Bulk Rename Utility”, which is very powerful in that way. After renaming the files, I then was able to search for them in Media Bay and quickly mass tag them. But, this of course has some issues associate with it: First, the license of some manufactures of sample libraries do permit that filenames are changed. Second, changing filenames can later prevent you from knowing where the samples originally came from. So to have a database maintained by an AI engine that tags the files based on content rather then a filename is actually the way to go, as it avoids both these issues.

With all respect I have to correct you on this one, yes every tag is written into the file in form of xml, you can open sample in text editor(look to be some short sample or you will wait a bit :joy:) scroll all way down and you will see all tags there it works also for some mp3 samples i have. Btw i suggest to remove read-only from all files before tagging.
“.steinberg”. Folder u mentioned i don’t remember i see something similar in my sample folder hmm maybe it’s hidden on system level but I am curious to check that again.

If they can just fix number ordering bug, I’ll be relatively happy, because that’s OCD seizure inducing.

I would love an auto tag system in media bay!!!
That would be a dream come true. I tagged about 200.000 sounds and loops during all these years and it was a big amount of time that I take off from composing.
Not sure that everyone knows but in cubase 11 there are new tags under musical category: articulations and moods.
Articulations is used for open, closed, muted etc… very useful for hi hats for example. But it’s just text and it’s not very handy. I’d like to have an interface like in ‘character’ where you can switch between options

Can anyone suggest an auto tagging app which is compatible with steinberg media bay tagging system? I tried some with no luck

1 Like

It can be done with a script. The mediabay.db can be opened with SQL lite. The script can check the bpm/key in the file name and updated with SQL statements, in the medabay.db database.

Example SQL:

Install sqlite and sqlite gui (easy to use to find field names)
Open sqlite3 (MAC)

change directory where mediabay3.db is located
sqlite3 mediabay3.db

make perl/python script that reads the filename
use that filename to find the id of that record in table media

sqlite> select id from media where FileName = “Lev_Shifted_029.wav”;
use that id to set MusicalTempo / key and other fields
sqlite> update media set MusicalTempo=120 where id=226306;

sqlite> select MusicalTempo from media where id=226306;
—nothing to display / not filled —
sqlite> update media set MusicalTempo=120 where id=226306;
sqlite> select MusicalTempo from media where id=226306;
120.0 (and the result / also shown within cubase mediabay)

This can probably be done in automator (mac).
If you rescan that directory all these (added) info will be deleted and you’ve to run your script again

1 Like

Edit: initially viewed on my mobile phone so just saw the first 2 lines of your post.

Logging in on desktop tonight i can see more detail - thanks!

Hi @stamb

Would you mind writing a step-by-step process for this?

1… Abc
2. …
3. Xyz

Maybe even with screenshots?

It think would help many people :muscle::pray::trophy:

A word of caution – Steinberg updates will not take modifications like this into account.

Unless you, yourself are a user of SQL databases and know how to write queries. I would be careful with this. In most cases the files have the data written to them, but there are a significant number of tags that are only written inside the mediabay database, and those can be lost forever when the db file is overwritten.

1 Like

@avviano; That Sononym looks super interesting. Is it replacing the mediabay entirely for you? (so you would drag files from Sononym into the Cubase timeline?)

Much of this thread is blowing my mind… the mediabay is deep but inconsistent to me. I actually think some tags are written to the files but there is also that large mediabay.db file that plays into it.

I’ve put some time into tagging but it’s a slow process. And some of the tagging (genre/descriptors?) don’t always seem appropriate so I’m not always consistent with my decisions, if you know what I mean.

I also don’t really understand how it’s automatically updating itself as I add folders. Sometimes I need to do a manual rescan, sometimes not.
PLUS… I’m finding some wavs/folders don’t show up in mediabay at all. I haven’t figured this out yet.

Which one are stored in .db and not in files itself? I am just curious.