Databending is the alteration of raw data to influence the data’s interpretation. Databending is commonly used to create in sound and image files effects that are exclusively digital in nature. The methods employed in databending are as varied as the results achieved.
Databending audio files
There are many various, and largely undefined, methods of databending audio files; some more involved than others. While there are programs available that will perform certain databending processes on audio files (such as granular synths), some of the simplest methods require nothing more than programs you already have installed on your computer.
Text Editor Bending
One of the easiest methods of databending audio files starts with opening the audio file in a text editor, such as Notepad. The text editor will display the raw data of the file, in text format. Databending at this level becomes as easy as selecting a chunk of text and altering it by deleting or copying and pasting the text, or replacing random characters within in the file. Unlike most databending technicues, this method tends to work better with MP3 files than with WAV files, which are less flexible. With WAV files, the most typical result is the entire WAV turned into a square chunk of white noise distortion.
All of the databending results in this tutorial are based on this recording (a performance culled from some track I recorded sometime earlier this year ). Databending is a destructive process. Always remember that you should never attempt to databend an original recording – only databend copies.
With this track, I have simply opened the file in Notepad and saved it under a new name. You can open an MP3 file in Notepad by right-clicking on the file itself, choosing “Open with…” from the menu options and selecting (or browsing to) Notepad from the Other Applications list; or by opening Notepad, and then from the File menu select Open, and then in Open dialog window change the file type from the default Text Documents (*.txt) to All Files (*.*). Once the file is opened in Notepad, go to File -> Save As and save the file under a new name, being sure to retain the .MP3 file extension.
If you’ve made it this far, you have officially databent a music file. What does this mean? Well, it means that the file is no longer a standard MP3 file, and that means that the file you have just created may be interpreted differently by different audio players (and even among different versions of the same media player). For example, the file above is rendered by SoundCloud as a slightly blurred and greatly abbreviated version of the original, but Windows Media Player renders it as warbly, slightly glitchy, but full length. See below for more on this. Once you have databent your file, try it out in various media players and editors to find out which gives you the most desired result. Then, capture that result by either saving a new copy of the file in a non-lossy format or making a new recording of the playback. For this, you may have to get creative, but that’s all part of the fun.
With this revision, I have opened the file in Notepad, selected a random chunk of code, copied it and then pasted it numerous times immediately after the initial selection. The trick to editing the raw data is PATIENCE. Unless you’re a computer, you’re not going to have much luck deciphering the endless string of gibberish that awaits you in Notepad, so there’s really no way of knowing how your edits are going to turn out, until after they’re completed. If your results aren’t particularly satisfying (and you’ve already listened to the file in as many audio players/editors available to you), go back and try again. It can become tedious, and Notepad is not efficient with files this large so the process can become time-consuming, but the results can sometimes be staggering. (More so than these examples indicate.)
To demonstrate how a databent file can be interpreted differently by different audio players or editors: This is the exact same file as the Notepad 1 example above. I recorded the audio output of my soundcard to capture how the file is rendered in Winamp. Compared to how SoundCloud interpreted the file, there’s an incredible difference.
This is the exact same file as the Notepad 2 example above, only opened in Cool Edit Pro 2 (which, if you’re old enough to remember flares and 8 tracks, was the precursor to Adobe Audition). In the examples above, SoundCloud didn’t render much noticeable difference between Notepad 1 and Notepad 2 (if you pay super-fine, close attention, you’ll notice that Notepad 2 is ever-so-slightly longer), but in these examples,you can hear how differently a WAV editor interprets the data, versus Winamp, a media player.
(The silence at the beginning of the track is an artifact of the databending process. If I were incorporating this file into a production, I would edit something like this out, but for the sake of maintaining transparency and integrity in this tutorial, I am not editing these files.)
In this example, I did a little self-serving databending. For this one, I copied the text from the Controlled Dissonance home page and randomly inserted it into the raw data of the MP3. This first rendering took place in Cool Edit Pro 2. CEP didn’t really care for the new text, as is evidenced by the huge chunk of silence in the middle of the file.
This rendering is of the same file, only in Foobar 2000, instead of CEP. Media players generally have a better time interpreting MP3s that have randomly inserted text. The fun about this method is that you can use any text. Ever wondered how a passage from your favorite book might affect/effect your favorite song? Now you can find out. The possibilities are limited only by the boundaries of your imagination.
Image Editor Bending
If you have read any tutorials on databending images using audio editors, this is the inverse of that process. Instead of opening an image file in a WAV editor, we’re going to open a WAV in a graphics editor; specifically in Photoshop. Now, I understand that not everyone has access to Photoshop, but if you have a copy or are savvy enough to procure a copy, it’s a fantastic tool for databending. If not, try using what you have or can get. I haven’t tried this out on every graphic editor out there, so you might find some gems that I have yet to discover.
Once you have chosen the program you would like to use, you will need to open your audio file. Your image editing program should have an “Open as…” feature, or open a file in a RAW format option, that will allow you to open any file. In Photoshop you can do this in the File menu by selecting “Open as…” and then in the open dialog window change the file format to Photoshop Raw (*.RAW). Browse to your WAV file (remember that WAVs work best in this process, counter to editing in Notepad) and open. You’ll be given a window with some RAW format options. I would recommend playing around with these settings to find what gives you the best results.* You should end up with something that looks like this:
“Great. So now what?” Go to town. You can do anything you want to this file. When you’re finished, select “Save As” from the File menu and save the file, being sure to replace the default .PSD file extension with .WAV. Now, the file that you have just saved is not likely to open in a standard media player. You’re probably going to need a WAV editor to interpret the raw data. In my case, I’m going back to my trusty Cool Edit Pro 2 where, upon opening the file, I am prompted to define how the sample format should be interpreted. I typically choose 16 bit, 44.1 kHz, stereo but, as you can hear below, choosing different options will result in different audio files.
This was effected using only the Box Blur filter in Photoshop. When opened, I chose to have the data interpreted as a 16-bit stereo WAV at 44.1 kHz.
This is the exact same file as the example above, only I have chosen to have it interpreted as an 8-bit stereo WAV at 44.1 kHz.
Then, to quote my son, I went “a little bit crazy”. First, I opened the file in Photoshop. Then I enlarged the canvas size, giving the original image a nice, chunky, black border. Next I copied the original image into a new layer, so that I could effect that, without affecting the border. Then I used the Crystallize filter with a Cell Size setting of 67. Just for kicks, I lower the opacity of that layer to 50% and then flattened the image, before saving it as a .WAV file.
Interestingly enough, you can see the channel separation that stallio discusses in his guides for databending images (see links below). Careful, though, as these examples are LOUD.
The 16-bit version.
The 8-bit version.
Finally, I backtracked to the point of the image where I had just expanded the canvas size and selected a color range that highlighted everything but the darkest pixels, and then I deleted the highlighted area. After which, I saved the file.
As you can imagine, I’ve only just scratched the surface of what’s possible here. My goal is not to provide you with an utterly inclusive guide to databending (that undertaking would be a life’s work), but to offer suggestions that will hopefully inspire you to find new and creative ways to punch your binary in the babymaker and twist it according to your own predilections.
* You may get warnings or error messages if the image size you select will render an image that is larger or smaller than the size of the file you’ve opened. If this happens, just keep adjusting the size up or down until you hit on a size that works.
So, what about turning non-audio files into audio files? If you have a WAV editor that allows you to open RAW files, then this is a simple task. Simply select Open or Open As from the File menu, be sure to change the File Type option to All Files (*.*) and browse to a file. If prompted to select the format for file interpretation, try your options and find out what works best for the file.
Databending a CD
While not truly altering the raw data of a CD, the intent is to create similar unpredictable glitchy effects, so I felt that it fit the category well enough to include here. Databending a CD is a simple process involving a physical degradation of the CD’s playing surface in an attempt to disrupt normal playback. This can be achieved through various different methods.
The best thing you can do to avoid total failure while bending CDs is to take the process in small steps. Whether you’re cutting, scratching or marking, apply the imperfections a little at a time and test the CD between applications. Avoid deep cuts and keep away from the edges of the CD. Remember that the CD still has to be readable in order to play. Experiment on CDs that will sound interesting but won’t be catastrophic should you ruin them completely. Also, just like with databent audio files, these bent CDs will provide different effects in different playback devices. If you have more than one CD player in your arsenal, try your CD in all of them, to find out if one gives you better results.
What does a bent CD sound like? Frankly, it sounds just like a CD that you’ve accidentally scratched, but controlling the process does give you the ability to create desirable glitches that can be used in a variety of non-conventional audio styles, such as IDM, plunderphonics and noise. My personal favorite is to bend a CD of classical, choral or vocal music, then drench it in reverb for an instant lush atmosphere. The following is a clip of a tenor singing his rendition of “Love Is A Many-Splendored Thing”, initially clean, then with enough reverb to smear the sound out into an interesting soundscape.
Databending image files
Databending image files consists of altering the raw data of an image to affect the image in some regard. These methods include converting image files to non-image files and then back again, or opening the image in a text editor and editing the data representation.
To quote from the WikiPedia article:
“Granular synthesis is a basic sound synthesis method that operates on the microsound time scale. It is often based on the same principles as sampling but often includes analog technology. The samples are not used directly however, they are split in small pieces of around 1 to 50 ms in length, or the synthesized sounds are very short. These small pieces are called grains. Multiple grains may be layered on top of each other all playing at different speed, phase, volume, and pitch.”
In the strictest sense of the term, granular synthesis is not a form of databending, in that the raw data is not intentionally edited to affect its interpretation. It does, however, offer a more analog approach to the concept of databending. Instead of hacking at the raw data, granular synthesis hacks at the interpreted data, creating new sounds by re-interpreting the data as it is chopped up and reorganized. The results are typically more usable than databent files, and the process is greatly expedited by a host of available programs. For the sheer audacity of the concept, I’ve included granular synthesis on this page.
Here are some applications that employ granular synthesis, either for waveform creation or audio manipulation:
Granulator (link bypasses site’s navigational frame – Hello 1996!)
Did you know that the very first Intelligent Machinery compilation project was a databending compilation? – Bending the Binary