You can download a copy of our manual in PDF form: here.
Introduction: Musical Semantics
During your adventures in music production you will have most likely used words like bright, flabby or fuzzy to describe a sound. Our brains are pretty good at choosing these words, take Mr. Obama here for example:
He listens to the wicked guitar line and comes up with some words to describe its timbre. He can then use these words to describe the sound to other people. He might say during his next speech “You know what? I really like a good fuzzy guitar tone”. Most people listening would then be able to formulate some kind of idea about how he likes his guitars to sound.
This mastery of language allows us to communicate with others about how we think things should sound. But what if Mr. Obama wants to make his guitar sound fuzzy but he doesn’t know how? If only he could give his recorded guitar track to a computer and say “Make it sound fuzzy”. The trouble is computers are not very good at ‘hearing’ the timbre of a sound.
A person (such as yourself) might take Mr. Obama’s guitar track and simply apply a distortion plug-in, tweaking the parameters to achieve that holy grail of fuzziness. A computer, on the other hand, is unfamiliar with the way that humans talk, so how does it know which parameters to tweak?
Here is where the SAFE project steps in. Our goal is to try and computationally define words like fuzzy, so your computer can easily set the parameters of an audio effects plug-in for you. In order to do this successfully, we need to collect lots of data that tells us about the kind of words that people use during music production, and which properties of an audio signal they are using them to describe… That’s where you come in.
The SAFE Plug-ins (SAFE stands for Semantic Audio Feature Extraction) are a series of DAW plug-ins that allow the user to provide timbral descriptions of the audio they are processing. The plug-in then analyses the audio and saves the anonymous data to our server. This data is collected from all users and analysed to give a general synopsis of the types of sound that a given descriptor is used for.
All this information can then be used to create a series of ‘semantic plug-in settings’. Users will be able to load plug-in settings by typing in descriptive words regarding the timbre of the sound being processed. The more people who upload descriptors to the server the more perceptually representative the downloaded plug-in settings will get.
It works just like any other EQ out there, the main difference is the group of buttons in the bottom right corner. These allow you to save and load ‘semantic’ plug-in settings.
Saving Plug-In Settings
Once you have used one of our plug-ins to get your audio sounding awesome, you can save the plug-in’s settings. This just involves the simple task of writing some descriptors in the text box and hitting save.
During the saving process, the plug-in has to analyse the audio you are working on. In order to do this the audio needs to be playing when you hit save. Once you hit save, the button will notify you that the plug-in is recording the features of your audio. During this time you should not change any of the plug-in parameters. If you do, the plug-in will tell you off and cancel the saving process.
The plug-in can be set to save either locally on your own machine or to our server. This is done using the save location button which looks like this:
The current visible image will determine where the data is saved to. Clicking on the button will toggle the mode. The mode will default to saving to the server if the plug-in is able to access the internet. If not it will save locally.
To aid in the analysis of the data there is also an option to enter additional information about the audio you are working on and yourself. You can do this by clicking the additional information button which looks like this:
This will present you with a data entry screen that looks like this:
This information will be used to help in creating genre/instrument specific plug-in settings.
Whilst the plug-ins should be very easy to use, there may still be a few problems. Here are a few issues that we are already aware of.
If you are using Reaper you may need to do some additional setup to get the plug-ins working properly. When adding plug-ins to a channel, right click on the plug-in name and select “Send all keyboard input to plug-in”.
If you’re using any software/services that are blocking your host’s connection to the internet such as a firewall or a network monitor (Little snitch etc…), you will need to add a rule allowing the plug-in to send/receive messages from/to our server (as this is where the semantics data is stored). This can be done by adding the following details to your configuration:
- Allow Incoming connections from 18.104.22.168
- Allow outgoing connections to 22.214.171.124
The SAFE project will never store or receive any personal or confidential information about you and by changing your settings to allow connections from our server, you will not affect the current configuration rules of your network monitor.
Have A Nice Day Now!
Enjoy using these plug-ins and please try not to write too many swear words in the descriptor boxes.