Transcoding 101 with Telestream, Wowza, Microsoft, and Rhozet

For those who have questions about how to effectively transcode their source video for online use, representatives of Telestream, Wowza Media Systems, Microsoft, and Rhozet came together in an online roundtable hosted by StreamingMedia.com. The event provided a great deal of practical advice for both those just starting out and those with some transcoding experience.

The roundtable kicked off with Kevin Louden, project manager for Episode at Telestream, who advised that before people can transcode into a streaming format, they need to decide which format is the best choice. The current battle is between H.264 and WebM video, he said, with H.264 enjoying more use. As for the resolution, he suggested going with a smaller size since many devices offer good upscaling nowadays. An MP4 H.264 AAC file at 640x360 will play on almost anything, he said.

For those who are looking to shave some time of their encoding sessions, Louden advised trying single-pass encoding, rather than double-pass. Using it takes about half time, so videos can be ready much faster.

Charlie Good, the chief technical officer of Wowza Media Systems, picked it up there and dug in a little deeper with encoding settings. The Baseline profile is the easiest to play, he said, and many mobile devices only support Baseline video.

Having the right keyframe interval is important, said Good, suggesting an interval of between one and five seconds. A shorter interval leads to quicker startup times, he said. Silverlight video works best with a two-second interval.

Delivering multibitrate video is important, he said, because you don't know what network conditions your viewers will experience or what device they'll be using. When creating multibitrate video, it's important that the keyframes be aligned, so that the player can switch between streams as needed.

Alex Zambelli, a technical evangelist for the Microsoft Media Platform, gave advice on encoding to Silverlight. Users can encode at up to 1080p. Smooth Streaming, Silverlight's version of adaptive bitrate delivery, is used by most people.

As for how many versions of a file a content owner should produce for Smooth Streaming, Zambelli suggested 12, saying any more would be overkill. Typically, he said, create 4 for standard definition streaming, 6 for 720p, and 8 for 1080p. Your top quality level should be the original file's resolution. For the lowest, don't go below 288x160. For bitrates below 300kbps, he said, it's better to half the frame rate than to lower the resolution.

Last up was Jon Robbins, a solutions manager with Rhozet, who explained that it's important to monitor your transcoding system once you have one in place, to make sure that what you think is happening is really happening. He also explained the steps in transcoding, explaining how the audio and video portions of a file get transformed and the whole package is then reassembled.

When shopping for a transcoder, he advised, format is key. Make sure it can decode what you have and encode the format you need. After that, look at video quality and speed.

When it came time for the question and answer section, viewers showed that they still had lots of basic questions about transcoding. One viewer wanted Louden to explain why he recommended single-pass encoding when double-pass produces better video. He explained that it's a good solution for organizations trying to squeeze every second out of their workflow and gain as much efficiency as possible.

Other viewers needed some transcoding terms explained. Robbins said that the mezzanine format is what the video archive is kept in, and is the master file that all other files should come from. Content owners will go back to that original when making new versions in the future.

Robbins also explained Baseline, Main, and High profiles. They exist for hardware makers, he said, to give them and easy way to say what level of video they support. The Baseline profile is less complex and is typically supported by phones. Main is more complex and is used by set-top boxes. The High profile uses the most processing power, so it's supported by desktop and notebook computers.

Interested readers can view the entire archived roundtable here.