Recording speech with a personal computer

This article started out as a guide on getting digital audio files into a personal computer, but has later evolved into a guide on how to record speech using a personal computer (Windows, Mac, or Linux). Because software and hardware change so rapidly, I am going to limit this discussion to the general principles that should apply relatively well to various hardware/software configurations and operating systems.

The workflow

The recording workflow is very simple. You need a microphone, a microphone pre-amplifier, an analog-to-digital (A/D) converter, a data conduit, a personal computer, audio recording software. Figure X shows a typical workflow involved in speech recording. There are many possible variations on the theme, but the fundamental concepts remain virtually the same. What does vary, is the degree of modularity of your recording chain. For example, you may have a truly minimalist setup with a microphone that has a pre-amplifier and A/D converter built-in, and connects to your PC via USB and sends digital audio data directly to your audio software. One such example is the Plantronics 655 headset that I reviewed on this web site. At the other end of the modularity continuum you can have a very sophisticated setup with each module of the recording chain being responsible for one type of function only (e.g., a microphone, a phantom power supply, a pre-amplifier, a mixer, a PCI sound card, etc.). I recommend that you try to strike a good balance between simplicity and sophistication, so that you can get the best results without unnecessary technological clutter.

System versus dedicated audio devices

Regardless of the operating system, you should be able to distinguish between system-level audio and application-level audio. If you have only one device (e.g., a laptop computer's built-in audio chip), these two levels conflate into one. If you have more than one device (highly recommended), you can set up on as your default system device, and the other(s) as your audio software dedicated device(s). In modern personal computers, many programs use audio devices. Your web browser, your email client, your calendar, your desktop gadgets, etc., use the system-default audio device to play and record audio. If you are serious about making high-quality recordings, you should probably use a dedicated device that only your recording software has access to. This simple idea is one of the reasons why turn-key hardware/software recording solutions work so well. I would like to encourage you to obtain a dedicated audio device for your personal computer and give your recording software sole access to it. Figure X shows Windows 7's recording devices panel. Note that the Realtek Stereo Mix is selected as a default recording devices for the operating system, while the USBPre device will be used solely by my recording application.

Figure X. Windows 7's Sound panel

Recording hardware

There are two alternative approaches to PC recording: (1) turn-key, and (2) à la carte. I have tried both and prefer (2). The reason is simple, I really enjoy the freedom to pick and choose, mix and match, the hardware and software to meet my specific needs and budget. Turn-key solutions tend to be more expensive, but the typically work really well. Digidesign Pro Tools is a family of excellent turn-key audio recording turn-key solutions that are used in recording studios all over the world. They also make a few entry-level, affordable products that you may find excellent. However, be advised that they may not work with any other software and/or hardware.

Conversely, you can design your own workflow that gives you more flexibility, though perhaps with a little bit of extra effort on your part. As long as your devices and software are standard-compliant (and most are), you should be able to put together an setup that is as reliable as a turn-key solution.

Recording software

If you buy a turn-key solution, your choice of recording software is made for you. Otherwise, you need to decide what audio recording and editing software will work best for your needs. If you like open-source software, Audacity is an excellent choice. It used to be unstable and quirky, but the recent releases have been first rate. Audacity will record more than one track of audio and comes with a perfectly adequately set of digital processing tools. Check out my tutorial on how to set up Audacity for speech recording. If Audacity doesn't work for you, there are other titles to choose from. On MS Windows, you have GoldWave, SoundForge, Wavelab, and Adobe Soundbooth, all of which work really well, but differ in important ways.

Input = output

Whatever your choice of hardware and software, you must remember one thing: your hardware input must equal your software output in terms of sample rate, bit depth, and channels. If your A/D converter is set up at 48,000 Hz, 24-bit, mono, your recording software must be set up to the same specifications. Otherwise, you risk degrading your signals by introducing artifacts, such as clicks, crackle, etc. Also, remember to disable as much unnecessary software and hardware processes as possible, e.g., anti-virus software, email checker, blue tooth radio, desktop gadgets, etc.

Prepare files for analysis

I am assuming that you record at least at 48,000, 24-bit. For most types of acoustic analysis, you don't need to work with files that big, especially if you have an older computer or use analysis software that cannot handle 24-bit files. In such a case, you should probably batch resample your files to 16,000 Hz, 16-bit. You can use Akustyk's batch converter, or any other batch converting tool, such as GoldWave. Do not be tempted to do any other signal processing on your files. Noise reduction, compression, EQ, may make your files sound better to your ears, but will inevitably change the sample values in unpredictable ways, thus compromising the accuracy of your analysis.