Robot Mastering

Stephen – have you any experience with these “online” mastering services? Here is one: (link removed) Apparently you can send in an MP3 file and they will plug it into an “algorithm” of some sort and then send it back to you. Have any of your customers ever used one? It seems that your file goes through some kind of program where perhaps a compressor and EQ is added but I’m not really sure. Anyway, if you have any thoughts I would love to hear them. Thanks again.

It’s funny you should ask that now, because an old online acquaintance of mine is one of the people behind that specific service you mentioned (I have removed the name and link, because I don’t want to single any one company out). This person recently got in touch with me and wanted me to try out their online mastering service, which I refer to as Robot Mastering.

As somebody who does mastering myself, as well as working with several very high end dedicated mastering engineers, I was very skeptical of any kind of automated “Robot” mastering, but I agreed to try it out and give him some feedback.

At that time, their site wasn’t open to the public yet, and I could only try the free mastering option, which doesn’t let you modify any settings, and only gives you an MP3 in return.  I decided to give it a quick try with a 24 bit wave file of a mix I just finished for a singer/songwriter type client.  This song was somewhere between acoustic and straight ahead pop/rock, and the client wasn’t interested in having his mix be as loud as possible.  When I did the mix for him, I sent him a reference file with some “quick” mastering applied with the Slate FG-X mastering processor, just so he could hear what the mix would sound like at closer to current loudness levels, but I didn’t go all out with the loudness.  The wave file I uploaded to the robot mastering service was the original mix file, which did not have any limiting on it, and had plenty of headroom.

When I got the MP3 back from the robot mastering service, I compared that MP3 to the reference MP3 I made with the Slate FG-X.  It wasn’t even close.  The robot mastered version was WAY too loud and too bright/harsh, and did NOT fit the style of music at all.  My FG-X version, which I spent maybe 2 minutes making (just a “quick” reference, and not meant to be the final master), was MUCH better sounding.

So, I sent my comments into the robot mastering guys, and they were forwarded to one of their programmers. They were quick in explaining that with their premium paid services you had much more control and could set some parameters to determine how loud you wanted it, and somehow be able to fine tune the overall sound to better fit the style of music.  Supposedly they have some smart algorithms that are supposed to figure out the style of music and set things accordingly, but it really didn’t work for me with the free version.  They offered to set me up with a premium account trial so I could try it out with some of the advanced controls, but I haven’t had time to try it yet and frankly am not really interested.

This same company also did some promotional posts on Facebook, and understandably got quite a few negative comments from audio professionals and musicians about this entire concept.

One of the things I asked them about was how they handle mastering for a full album. Originally, one of the primary functions of mastering was about making all the songs for an album work together, so that you don’t feel like you need to adjust your volume or tone controls from one song to the next.  Then you would get the songs sequenced in the order you want, adjust the spacing between the songs, and create final masters for CD, Vinyl, and/or tape. How would this robot online mastering service be able to handle an album and generate a full master image for CD or other formats? They assured me that they were working on that, and would have that service available in the future.

My main gripe with these types of automated processes is that no matter how much processing power you have, these algorithms are never going to be able to replace the trained ears of a mastering engineer, or even a moderately talented musician/engineer.  Sure, automated processes can compare the frequency spectrum and dynamics of songs, and apply the necessary processing to try to match one up against another. There are already plugins that allow you to do that yourself.. that can easily be done with any computer these days. However, what they are missing is the human element that listens to the context and emotional aspects of a song and adjusts the settings in the most appropriate way for that particular song. Music is more than just frequency and dynamic curves, it’s emotional. So far, I have not heard of any algorithms that can decipher any emotional elements of a song.

Right now, I don’t believe these robot mastering services are good enough to replace us mere humans! As computers and technology continue to advance, it may one day be possible, but I still think I would prefer to do things myself.

Of course, I’m not the target market for these types of services.  I have many years of professional recording, mixing, and mastering experience, so I certainly am not going to pay for a robot to do something that I could do better on my own.

They say that these services are for the musicians who may not be able to afford a professional mastering engineer for every song they put out. I can almost understand this market, and it might work for them as there are many musicians who really don’t know what they are doing when it comes to mastering.  However, I also know many mastering engineers who will master songs from indy artists for $25 to $50 per song. Most of these mastering engineers will cut deals for full albums. Although I’m not a dedicated mastering engineer, I have done a lot of mastering for many artists, and I just charge my normal hourly rate, and can master most projects in an afternoon.

So, who is it exactly that can’t afford to pay someone like myself for mastering? My guess is it’s the musicians with the home studios, who are recording and mixing everything themselves. If those artists have the skills to get really great sounding recordings and mixes at home by themselves, then they already have the skills they need to do some self-mastering that would be at least as good as the robot mastering services. If, however, they don’t have the skills or the gear to get great sounding recordings and mixes on their own, then ANY type of mastering is not going to be able to make their songs sound good. Mastering is the last stage in the process, and can’t work miracles on poor sounding recordings & mixes.  It’s also usually the cheapest part of the whole process if you are already working with professionals for the recording and mixing. Thus, I’m having trouble figuring out who would actually use these types of online robot mastering services.

The guy I talked to thought it might be good for someone like myself when I was too busy to do the mastering myself. But, as stated above, I can crank out a “quick” master in a few minutes that already sounded better than their robot service was able to do.  It took much longer for me to upload an uncompressed 24 bit wave file and wait for their service to do it’s thing and email me a download link. In that time, I could probably do a full mastering job on the same song.

I realize I’m being fairly harsh, but people pay may to be as critical as possible in order to make their songs sound as good as possible. I’m certainly a technology geek, and I’m interested in these types of robotic processes from a technology standpoint.  Perhaps this type of thing will be more useful to help home based artists who don’t have an accurate monitoring setup, as it could help them detect and fix problems in their mix that resulted from them simply not being able to hear their work properly. In that case, though, I see the more helpful thing being some sort of plugin that you could put on various tracks, as well as your mix buss, in your DAW, and it can help you shape the tone of each instrument to better work in your particular mix. There may already be some sort of algorithmic mixing software out there, as I’m sure I read something about it in the past, but it hasn’t become too popular yet. As stated above, I do know that there are already plugins that will help you match the frequency spectrum of  an audio file against something else.  At this time, I see that type of technology being much more useful for the home recording people as it could help them make adjustments in the mixing stage, and would also help them learn where they have problems in their recording or monitoring setups.

We are certainly getting closer to automating much more of the recording, mixing, and mastering processes, but I don’t think the “robots” are yet able to decipher the all important human/emotional elements yet to make the best decisions.

Posted in Ask MusicTECH!
2 comments on “Robot Mastering
  1. fireflyva says:

    Stephen – I sent in an MP3 file (because it was free) to be “mastered.” After receiving it and running it through a EQ graphic analyzer, I noticed that other than the song being greatly boosted in terms of volume, a lot of the peaks were cut off, the song was a lot more “brittle” but also had the bottom end boosted. A friend of mine made the comment that it now sounded like the mix was optimized for ear buds. It did have a lot of punch but the frequency balance was not better. Anyway, just some thoughts and observations.

  2. MusicTECH says:

    Pretty much my same experience. To the untrained ear “louder” sounds better, as does applying a bit of loudness curve EQ (boost lows and highs, which again is making it louder). Those who know better are not so easily fooled. Or, perhaps the reference curves they are comparing against are some of the commercial examples of the extreme loudness wars. For the musicians who don’t know better, they might get these “masters” back and think it sounds better simply because it is loud… maybe that’s their target market. I don’t necessarily think their idea is all bad, but at least their free mastering algorithm isn’t good enough yet.