The Mysteries of Databending Explained

Posted on April 08, 2016 by Skylar Trainor

Any digital file, be it a video file, an image file, a sound file, or an executable, exists as a mathematical construct; a collection of numbers. I'll be the first to admit that the mathematics underpinning these files is a foreign language to me. However, to Databend a file you don't need to know or understand any of the esoteric mathematical voodoo underpinning the program. It might help you to research it but it isn't necessary to know if you just want to generate random sounds. To put it simply there is a lot of art and very little science involved in this technique.

Databending is very similar to Circuit Bending. You are taking something meant for one purpose and using it in the context of another. With Circuit Bending the classic original example that everyone uses is taking apart a speak and spell and rerouting the internal circuitry to make the speak and spell spit out random gibberish. You can even get really fancy and technical with it and add potentiometers and switches to increase your control over the electrical current. Here's an example of databendending in action (from our Databenders_Toolkit release)

In contrast, Databending doesn't involve any sort of physical analog audio or electrical circuitry. Its all digital, it all happens on your computer. They get compared a lot because the resulting sounds and images are often chaotic and unpredictable.      Databending is actually cheaper, easier, and less technical then circuit bending. You don't need a thirty dollar soldering iron, wires, or a overpriced vintage children's toy. You just need Audacity, which is a free digital audio workstation.           

Databending is the art of digitally manipulating a file meant for one format in a editing program meant for a file of a different format. So you could open an image file as a text file in notepad and delete a paragraph. That would alter the image. You could open that same image in audacity as a wav file and apply a delay then export it. That would also alter the image. But I never use image files because usually don't yield cool sounds. I use executable files.

Executable files, especially older games, are often very large and complex. So logically you would have a higher chance of generating something cool if you have a larger pool of raw data too feed into audacity. And when I say raw data I mean literally raw data. That's the function you're going to use in Audacity to generate these sounds.

The number one mistake people make when they're trying to databend is using small or heavily compressed files. The bigger and more complex the file the more sound it will yield when you bend it. Executable files work great. Also don't be afraid to play around with the Byte order.

This is the most effective way I've found to get cool sounds with this technique. Its really easy and really quick. The catch is that it often spits out painful random noise which isn't really useful without editing and post processing. Like this:

So turn your volume down, and I mean way down, when you are working with this technique. What I often do is after I generate a wav file I go to the effect section and use the Amplify function to decrease the decibel level by around minus fifteen to minus twenty decibels. No sound design project is worth getting tinnitus.

So once we have turned our wav file's volume down we can move on to editing. I look for little tonal bleeps and interesting sonic artifacts in the wav and start breaking it down into its component parts before I apply any effects. Most of the files I deal with when data bending are ninety percent white noise and ten percent dying robot sex sounds. They do say beauty resides in the eye of the beholder, or in this case the ear of the beholder.

             For Example:


On to post processing:

The effects I most often use now for mixing fixing issues in my sounds are time stretching, EQ, multi-band compression, and reverb. These Databent sounds often have imbalanced highs and extreme low ends so those get cut first and the energy gets reassigned to the sub bass region.

Because we lowered the gain of our wav file in the last step we have some headroom to play around with. So we have the choice between using additive and subtractive equalization without having to worry about clipping.

In the post processing example I used a really tight reverb with a long decay. It was just the stock FL studio fruity reverb. If you really want to recreate the exact sound of the Databenders_Toolkit pack you would need an actual tape machine and whatever hard compressor(s) equalizers and saturators Teddy used. It would probably be an expensive venture to copy his studio set up.

You would also need Reason since that's what I used to process the original pack. Reason is a fun tool. Its basically a giant modular synthesizer/studio that models control voltages in a similar way to what you would see in a actual studio. Sound quality isn't as good as real hardware but it nails the feel.

I did this pack many months ago so I don't have all the patches I used. One technique I do remember using was running the waves through Reason's vocoder and automating the individual bands so it would cycle through the frequency spectrum. That then went through a reverb.

Some of the other sounds were time-stretched using Paul Stretch and the Akaizer. These are both free effects you can download on the internet. I know one went through an additive EQ with the boost frequency being modulated by a control voltage signal.          

So that's the process in a nutshell. To recap my points: Don't use small compressed files. Turn your volume down before you playback or eq your files. Check your frequencies paying extra attention to the 0-40hz, 220-400hz, and 2-5 khz ranges.



Check out my Soundcloud. I'm pretty active on it and its the best way to get in touch with me.

Posted in Databending