Most seasoned compressionists know that you can’t have high-quality streaming media without quality audio and video. In honor of our production issue, I thought I would list the production techniques that can make or break audio and video quality.
Let’s start with the setup. There are two concepts to keep in mind here: contrast ratio and detail. Cameras have limited contrast ratios (the ability to retain detail in both the lightest and darkest regions in a frame); codecs have even less. If your background is bright white and your subject is wearing a black suit, your suit will likely look like a black blob with no detail, which will really impress all those Millennials watching low bandwidth streams on their cellphones.
Codecs also have limited ability to retain data. When you pack a background with excessive detail such as venetian blinds or finely patterned wallpaper; dress your subjects in madras or thin pinstripes; or include motion such as windblown grass, shrubs, and trees in the picture, you stress the codec’s ability to retain good quality throughout the frame. The codec can’t tell if it’s the subject’s face that’s important or the leaves or grass behind her, so it wastes valuable bandwidth preserving detail in the grass, making the face blocky. The unfortunate reality is that backgrounds that look ideal for television can be awful for streaming.
The shame is, until about 3 or 4 years ago, streaming bandwidths were sufficiently high to retain very good quality irrespective of detail or motion. Then mobile arrived, and data rates dropped dramatically. If many of your viewers are watching on mobile devices, you have to make sure the video looks good at 640×360 @ 600Kbps in addition to 5Mbps at 1080p. This means eliminating unnecessary detail and motion.
Once you have the set and clothing nailed, lighting is the next major production issue. One common mistake is to mix lights of different color temperature, such as incandescent and fluorescent lights, though the decline of incandescent lighting in general has made this less common. But many producers still use inadequate lighting for shoots, which not only mutes your colors but introduces grain into your video, which is more detail that the codec has to preserve, which means lower quality.
From my perspective, unless the subjects are complaining, you probably don’t have enough light, though that’s another problem inexpensive CFL and LED lighting has reduced. From a technical standpoint, you should throw enough light on the subject so that your zebra stripes set to 75 IRE light up the subject’s face with shutter speed at 60 or higher and zero gain. If any of these terms are unfamiliar to you, Google them and learn what they mean — setting exposure correctly might be the most important skill for any streaming shooter.
While you’re on Google, search for “rule of thirds” positioning and the technical definitions of long shots, medium shots, and close-ups. Nothing screams amateur more than video that’s improperly framed; pros will immediately see it and wince, while less technical viewers will just have a vague uncomfortable feeling that something is wrong with the picture.
Camera quality matters, even for 320×240 picture-in-picture videos. I recently compared three webcams and two camcorders — including a $3,000 prosumer camcorder — for close-up webinar production. Even in a small tiny viewing window, the prosumer camcorder produced a subtly better image than the $900 consumer camcorder, such as how a bespoke suit compares to an off-the-rack “bargain.” Not surprisingly, both camcorders completely outclassed the webcams.
Finally, audio is at least as important as video, perhaps even more so. Always, always, always use an off-camera microphone, even when standing next to the camera. During all events, monitor audio levels during capture. Unless you have someone dedicated to audio capture, consider using automatic gain control, which usually does a better job than a human can do. Never ship VOD files before normalizing the audio to a level at or near 0 dB, and consider applying compression to make any spoken words more distinct.