RealNetworks pioneered streaming audio with RealAudio, the first streaming media product for the Internet. Since its debut in 1995, RealAudio has become the standard for network audio, delivering stereo sound over 28.8 Kbps modems and CD-quality sound at high connection speeds. This chapter provides a reference for the RealAudio codecs, and gives pointers on how to prepare and encode your sound files for streaming or downloading.
 |
For More Information: See also "Adjusting Audio Gain" and "Audio Delay Compensation Prefilter" for information about audio filters that you can use with RealProducer. |
Because RealAudio clips are compressed, you typically start with a sound file
in a digitized, uncompressed format such as WAV or AIFF. Using
RealProducer, you create a RealAudio clip from the source file. RealAudio
clips typically use the file extension .rm, although clips may also end with
.rmvb (variable bit-rate clip) or .ra (audio file created by RealPlayer). The
following sections explain how RealAudio encodes an audio file for streaming.
This information will help you to produce high-quality streaming clips.
One way that RealAudio codecs squeeze an audio file's size down is by throwing out nonessential data. This makes it a lossy compression format. RealAudio doesn't delete data indiscriminately, though. It first jettisons portions you cannot hear, such as very high and very low frequencies. Next, it removes as much data as needed while keeping certain frequencies intact. Voice encoding favors frequencies in the normal human speaking range. Music encoding retains a broader frequency range.
Although RealAudio is savvy about what audio data it throws out, be aware that the lower the target streaming speed, the more data gets ejected, and the cruder the sound quality becomes. At low bandwidths, you get roughly the quality of an AM radio broadcast. With faster connections, you can encode music with FM-quality sound. And at the high speeds of DSL, cable modems, and LANs, RealAudio sound quality rivals that of CD or multichannel DVD playback. When creating RealAudio clips for any bandwidth, it's important to start with high-quality input, as described in "Audio Capture", to attain good sound quality.
You create a RealAudio clip by using one or more RealAudio codecs. A codec is a coder/decoder. It tells RealProducer how to turn audio source files into RealAudio clips. On the receiving end, RealPlayer uses codecs to expand clips into audio data the computer can play. RealAudio employs a series of codecs, each of which creates an audio stream for a precise bandwidth. One codec compresses mono music for a 28.8 Kbps modem. Another one compresses stereo music for that same modem speed. This set of codecs is different from the set used to compress music for, say, DSL and cable modem connections.
A RealAudio clip consumes bandwidth at a flat rate determined by the codec used to encode the clip. A RealAudio clip encoded with a 20 Kbps codec, for example, steadily consumes 20 Kbps of bandwidth as it plays. The following table lists the standard bit rates for RealAudio clips encoded for specific target audiences by RealProducer. Encoding a voice-only audio file for a 28.8 Kbps modem, for example, creates a 16 Kbps streaming clip. With mono music input, though, you get a 20 Kbps clip.
 |
Tip:
In terms of bandwidth use, RealAudio is inflexible. The
RealAudio codecs set streaming bit rates in a stairstep model:
20 Kbps, 36 Kbps, 44 Kbps, and so on, with no inbetween
choices. Because RealAudio clips always stream at specific bit
rates, consider their bandwidth needs first when you use them
in multiclip presentations. Then create your other clips to
stream within the remaining bandwidth.
|
|
For More Information: With SureStream technology, a single RealAudio clip can stream at many different speed. For the basics of SureStream, see "SureStream CBR Clips". |
When you encode a RealAudio or RealVideo clip using RealProducer, you do not select the RealAudio codec directly. Instead, you choose your audience settings, as described in the section "Choosing Audiences". You might select one audience, such as 256 Kbps DSL users, or encode multiple audiences into a single SureStream clip that can stream at different bit rates.
Unless you wish to change the audience defaults, you just need to select the right audience settings when you encode a clip to have RealProducer use the appropriate RealAudio codec. The audience templates presented in the RealProducer graphical user interface always indicate the target streaming speed, such as with the "384k DSL or Cable Modem" audience.
Note, though, that the type of audio input you use can affect your choice of audience. If your are using stereo surround audio as your source, you may want to encode using the "350k Surround Stereo" audience template instead of a template that encodes just standard, two-channel stereo audio. Or, if the input is multichannel, you may want to choose the "350k Multichannel" audience setting. The following sections explain more about these audio types.
Additionally, you should note RealPlayer compatibility with the RealAudio codecs you use. Old versions of RealPlayer can play audio encoded with voice codecs, for example, but only RealOne Player and later can play audio encoded with certain stereo, stereo surround, and multichannel codecs. The sections that list the available codecs explain compatibility issues. Note, too, that all RealPlayers since 1998 have the ability to upgrade automatically to new audio codecs or a new player version if the user attempts to play an unsupported RealAudio format.
|
Tip: Chapter 9 explains how to determine, and change if necessary, the codec settings used with each audience setting. If you wish to change a default setting, be sure that you understand the codec properties described in the following sections. |
The following sections describe the RealAudio codecs available through RealProducer. The codecs are listed in separate tables for voice, mono music, stereo music, stereo surround, multichannel, and lossless audio. Each table provides the following information.
The Codec column describes the codec as it appears in the RealProducer
interface. The name lists the streaming bit rate for the encoded audio, and
indicates the type of audio input the codec is suited for.
The Type and Flavor columns identify the codec for use with the command-line
application. This information is not shown in the RealProducer graphical user
interface.
The sampling rate column lists the codec's optimum sample rate. If the input does not have the expected rate, RealProducer resamples the input without causing pitch shifting. For best results, the audio input should have the same or a higher sampling rate than the codec's expected rate. For SureStream clips, the input sampling rate should be equal to or greater than the largest sample rate of all codecs used. Although RealProducer accepts audio input at any sampling rate, it is optimized for the following rates:
|
Tip: After you load a digitized clip as input, a process described in "Using a File as the Input", you can click the Source Properties button to display the audio properties, including the sampling rate. |
|
For More Information: RealProducer allows you to choose high- quality resampling (recommended) or fast resampling. The section "Setting Audio Parameters" explains how to set this option in the graphical application. See "Audio Resampling Quality (-arq)" and "Media Profile Properties" for information about setting this option through the command-line application or job file, respectively. |
Voice codecs produce the best results for voice-only audio input. The lowest- speed voice codec normally used to encode a RealAudio clip streams data at 16 Kbps. The lower-speed codecs (5, 6.5, and 8.5 Kbps) are used as SureStream duress streams when the connection bandwidth drops. They're also used to encode soundtracks for low-bandwidth RealVideo clips. The following table lists the available voice codecs, which are compatible with RealPlayer G2 and later.
Music codecs are designed to encode audio with a larger pitch variance than voice. You will capture a broader, fuller sound with codecs designed for higher bit rates. As with the voice codecs, the lowest-speed mono music codec normally used with RealAudio streams data at 16 Kbps. The lower-speed codecs (6, 8, and 11 Kbps) are used as duress streams in SureStream clips, and to encode soundtracks for low-bandwidth RealVideo clips. When there are two versions of a codec, RealProducer uses the higher-response version by default.
The 20 kbps, 32 kbps, and 44 kbps music codecs come in two varieties. By default, RealProducer uses the "high response" versions, which are the better codecs for most situations. But you can also use the "normal response" versions by changing your audience templates, as described in Chapter 9.
The high response codecs cover a larger frequency spectrum than the normal response versions. Sometimes, the high response version has twice the range as the normal codec. This means it provides crisper sound and is better at capturing high frequencies. With symphonic music, for example, the high response codec gets more of the flute and piccolo. It can produce more distortion than the normal response codec with voices and loud sounds such as drums, though.
If you are encoding music with a diverse range of frequencies, stick with the high response codecs. If you notice distortion, compare your results with a clip that uses the normal response codecs. The best tool for determining which codec to use is your ear. Listen carefully for minute differences in how the clip sounds. It also helps to have other people listen. Our own ears have different frequency responses, too.
You can use the following codecs to encode audio files as mono music. All mono music codecs are compatible with RealPlayer G2 and later.
Use stereo music codecs for encoding traditional, two-channel stereo music. RealProducer also uses these codecs when you encode voice-with-music clips.
You can encode many different bandwidths of stereo music, using three different stereo codecs:
 |
Note: At several bandwidths, you can choose between normal and high response versions of the stereo codecs. RealProducer uses the higher-response version by default. But you can change to a normal response version as described in "Creating and Editing Audiences". |
|
For More Information: The section "About High-Response Codecs" explains the difference between normal and high- response audio encoding. |
The following stereo music codecs are available in RealProducer 10.
Encode your audio using a stereo surround codec if you know that the source audio is matrixed, multiple-channel sound, and you wish to preserve the multiple channels for your listeners. Because stereo surround is compatible with conventional stereo systems, listeners who do not have stereo surround- enabled equipment will still be able to hear the two main channels. Stereo surround audio and video clips are suitable for streaming or download.
Stereo surround audio includes more channels than traditional stereo, which uses just the left and right channels. These additional channels are mixed (or matrixed) into the conventional left and right stereo channels. This allows older receivers to play just the left and right channels, while newer receivers enabled for stereo surround can extract from the left and right channels the audio data for the additional speakers.
|
For More Information: You can find background about producing stereo surround input at the Dolby Laboratories Web site at http://www.dolby.com/tech/. See also http://www.realnetworks.com/resources/index.html for additional information. |
The RealAudio stereo surround codecs preserve the matrixed, multichannel surround audio in the audio input. RealProducer supports any number of matrixed channels. Because the audio input is standard stereo input, the computer running RealProducer does not require a special sound card or cabling. The following table explains the common channel arrangements found in matrixed, stereo surround audio.
It is important to note that RealProducer does not mix the multichannel stereo surround into the left and right stereo channels itself. The source you are encoding, whether a static clip or live input, must be matrixed already. This type of audio content is typically created using encoders by Dolby (http://www.dolby.com), CRS Labs, or Digital Theater Systems (http://www.dtsonline.com/). These sources are prevalent on DVDs and television broadcasts. Although you can use non-digital multichannel sources, digital sources provide the best results.
To hear the matrixed, multichannel sound, RealPlayer users can play the audio on an A/V receiver, such as a home theater system, that is equipped with stereo surround decoding, and that is connected to the surround channels and optional subwoofer. As well, some computer speakers will play stereo surround audio directly. Audio systems not enabled for stereo surround play just the standard left and right stereo channels.
If you do not want to preserve the stereo surround information, you can encode your audio with an audience template that uses the conventional stereo codecs described in "Stereo Music Codecs". The primary reason to do this is to stream at bandwidths lower than those available for stereo surround. By using a SureStream clip, you can encode low-bandwidth streams in conventional stereo, and high-bandwidth streams that use stereo surround audio.
If your audio source is traditional, two-channel stereo, do not encode the input using stereo surround audio codecs. Although traditional stereo encodes OK as stereo surround audio, RealProducer does not create the extra channels (it only preserves existing channels), so using stereo surround audio codecs does not enhance the listening experience. Additionally, the standard stereo codecs are more efficient at encoding two-channel stereo than the stereo surround audio codecs.
RealProducer can encode stereo surround audio with any of the following codecs. The three "RealAudio" codecs are compatible with RealPlayer 8 and later. The "RealAudio 10" codecs, which are based on AAC technology, are compatible with RealOne Player (a codec autoupdate is required) and later, including RealPlayer 10.
The multichannel RealAudio codecs preserve the discrete, multiple channels in the audio source. Use them if you know that the source audio includes multichannel sound, and your intended listeners have home theater systems or other equipment able play all of the channels. Multichannel audio and video clips are suitable for streaming or download.
Like stereo surround audio, multichannel audio includes channels in addition to the left and right stereo channels. Unlike stereo surround, however, multichannel audio encodes the additional channels separately, rather than mixing all channels into the signals for the left and right speakers. For this reason, multichannel audio is often called discrete multichannel, rather than the matrixed multichannel of stereo surround audio.
Using multichannel audio preserves the optimum sound quality of multichannel audio. With stereo surround, the matrixing process may replicate some audio data meant for one channel on another channel. (This is an artifact of the stereo surround mixing in the audio source, rather than the RealAudio encoding.)
To use discrete, multichannel audio, your sound system must capture and preserve each channel. If you start with an uncompressed, prerecorded file, for example, that digitized file format must preserve the additional channels. A common audio format to use with multichannel audio is AC3, which can be digitized as an MPEG, QuickTime, AVI, or Wave file. The sound card used with the RealProducer computer must also support the additional input channels. A standard sound card supporting only two-channel stereo input will therefore not work for discrete, multichannel audio.
RealProducer encodes all multichannel output as 5.1 channels (left, center, right, left surround, right surround, LFE bass). It can accept fewer than six channels as input, upsampling as necessary to create the 5.1 channels. The obsolete quadraphonic multichannel format, which uses two front and two back channels, is not supported.
RealProducer can encode multichannel input from a file but not from a live capture. You can use the following audio source formats as input for the RealAudio multichannel codecs:
.wav).avi).mov)To hear the different channels in discrete, multichannel audio, RealPlayer users on Windows can direct the audio to a multichannel-enabled sound card or home theater system. With traditional speaker systems, and on operating systems other than Windows, RealPlayer converts the audio signal to the standard stereo channels.
As with stereo surround audio, you can encode multichannel audio with standard stereo codecs. This does not preserve the multiple channels, however. For example, you might create a SureStream RealAudio clip that streams multichannel audio at high bandwidths and standard stereo at low bit rates. You should not encode standard stereo input with the multichannel codecs, however, as the quality will not be as high as when you use standard stereo codecs.
The following codecs are available for high-bandwidth, multichannel recordings. All multichannel codecs are compatible with RealOne Player (a codec autoupdate is required) and later, including RealPlayer 10.
RealProducer 10 includes a lossless RealAudio codec that faithfully
reproduces the full dynamic frequency of the input audio file while
compressing the output. The encoded clip, which is saved in the RealMedia
variable bit rate format (.rmvb), is typically around half the size of the input
file, though the compression rate varies with different types of input.
The RealAudio lossless codec is designed primarily for high-quality music downloads in mono or two-channel stereo format (multichannel output is not supported). It replicates CD-quality sound in a format that takes less time for the user to download.
The RealAudio lossless codec preserves the exact sound of the input audio while compressing the file size to about half the original size. Although the lossless audio codec is designed for high-fidelity music downloads, you can also use it with with streaming clips and broadcasts in high-bandwidth environments.
|
Note: The lossless codec is not accessible through the graphical user interface. Use the command-line application as described in Chapter 14 to encode your files as lossless streams. |
You can encode any audio or video format acceptable to RealProducer 10 with the RealAudio Lossless codec. The codec is optimized for the 16-bit, 44.1 KHz audio used in audio CDs, but other sampling rates and bit depths are accepted as well. During encoding, you can apply the audio delay, audio gain, or audio meter prefilter.
The primary, intended output for lossless encoding is a single audio stream saved as a downloadable clip. However, you can also use lossless audio when encoding constant bit rate or variable bit rate video clips. You can also use lossless audio encoding for streaming clips or broadcasts, combining the lossless audio track with other audio tracks in a single SureStream clip or live stream.
Lossless audio clips can be streamed and downloaded for playback by RealOne Player (an automatic codec update is required) and RealPlayer 10. You can stream or broadcast using lossless audio with Helix Server version 9 and later servers. The legacy push broadcast mode is not available for broadcasts using lossless encoding.
Although a lossless audio stream is intended for downloading, it is also designed for streaming. However, unlike audio streams encoded with other codecs, a lossless stream does not have a single target stream rate. Instead, the streaming rate is approximately the size of the audio data in Kilobits divided by the audio duration in seconds. For example, a lossless audio clip that is 6 Megabytes (49,152 Kilobits) and plays for 2 minutes (120 seconds) streams at a rate of 410 Kilobits per second. Use of lossless audio for streaming media is therefore recommended only for high-bandwidth situations.
|
Tip: If you use lossless audio in a streaming video clip, keep in mind that the visual track is compressed to fit within the remaining bandwidth of the overall target rate. If the audio uses up most of the streaming bandwidth, the video's visual quality will suffer. |
RealProducer 10 provides three encoding modes—low, medium, and high—that
affect video and lossless audio. Both modes faithfully reproduce the full audio
range of the input file. The higher compression modes perform more complex
analysis on the input, however, resulting in longer encoding times but smaller
file sizes. Because of the reduced file sizes, the higher compression modes
create lossless audio clips that stream at lower bit rates.
With one exception, a lossless audio clip cannot be modified:
.wav file, or to a lower-fidelity format (such as MP3) for transfer to portable devices.Use the following lossless codec to encode perfect quality sound. The RealAudio lossless codec is compatible with RealOne Player (a codec autoupdate is required) and RealPlayer 10.
| Codec | Type | Flavor | Sampling Rate |
|---|---|---|---|
| RealAudio Lossless Audio | ralf | 0 | 44.1 kHz |
A streaming clip reflects the quality of its audio source. Any quality problems within the source will affect the streaming clip as well. Because you cannot edit a broadcast, live Webcasting introduces several issues beyond those involved with delivering on-demand clips. This section will help you capture high-quality audio source files, or set up your sound equipment to deliver good broadcasts.
|
For More Information: For information about broadcasting live content, refer to Chapter 10. |
If you plan to stream existing material, start with the best source possible. Use the cleanest recording with the least amount of unwanted noise. Compact discs (CDs) and digital audio tapes (DATs) are good source media, although well-recorded analog sources such as records, reel-to-reel tapes, and chrome (type II) cassettes can sound just as good. Try to avoid consumer-grade recording media such as Type I cassettes and VHS tapes.
Every piece of equipment in the audio chain—microphone, mixer, sound card, and so on—affects sound quality. If you intend to provide professional-quality audio content, invest in professional-quality audio equipment and software. Lesser equipment can add hiss and distortion, degrading sound clarity.
It is important to use high-quality, shielded cables. Using unshielded cables increases the likelihood of introducing line noise and radio frequency interference into recordings. Keep audio cables physically separated from power cords to minimize the introduction of noise. Also be sure to ground all equipment properly.
Setting correct input levels is crucial. All audio equipment has a signal-to- noise ratio, the ratio between the loudest possible sound the equipment can reproduce without distortion and its inherent "noise floor." Also called "clipping," distortion of this type is audible as a high-frequency crackling noise.
To get the best signal-to-noise ratio, set the input level on each audio device in the signal chain so that it uses its full range of available amplitude without distortion during the program's loudest sections. The signal chain typically includes a microphone, a mixing desk, a compressor, and a sound card. For each piece of equipment, set levels as close as possible to 0 decibels without going over that level.
Check for signal distortion at each point in the signal chain. Perform several test runs, and make sure that there are no peaks above maximum amplitude. Adjust the levels on your sound card mixer so that the input approaches but does not exceed the maximum. Be conservative, though. Levels might suddenly increase if, for instance, an interviewee suddenly speaks loudly or a crowd at a sports event roars.
When broadcasting live audio streams, it is useful to have a dynamics compressor for gain compression (not data compression). This piece of audio equipment automatically adjusts the volume level. By providing a consistent volume level, it allows you to "set and forget" the input levels to RealProducer.
Try to capture sound with a sampling width of 16 bits. RealAudio codecs have different sampling rates that produce the best sound, however. If your sound card allows it, capture audio content at the optimum sampling rate for the codec you intend to use. The RealAudio encoder will convert the file to the optimum rate if necessary, but this is recommended only for static files. For live broadcasts, use a sound card that supports the optimum rate. This avoids the overhead entailed in converting the rate while encoding sound in real time.
|
For More Information: "RealAudio Codecs" lists the optimum sampling rates for each codec. |
|
Tip: You do not need to capture stereo sound if you plan to use a mono codec. However, many sound cards simply discard the right input channel in mono mode. If you have a mixing desk, pan all inputs to the center so that nothing is lost during the conversion to mono. |
If you are not broadcasting audio live, you work with digitized audio source files in supported formats such as WAV or AIFF. You then edit the audio files to optimize them. To do this, you need to be familiar with the features your editing program offers. This section gives you some optimization tips you can try with your editing software before encoding your clips with RealProducer.
|
Tip: Always keep copies of your audio source files. You cannot convert RealAudio clips back to their original source formats. |
DC offset is low-frequency, inaudible noise that results from equipment grounding problems. If you don't remove it, it can skew the results of subsequent sound editing. Use your sound editor's DC Offset function immediately after recording a digital audio file.
|
Tip: If your editing program has this option, remove DC offset during recording. This eliminates an editing step. |
Set sensible input levels when recording, and then use normalization to maximize the levels after recording. Your streaming files sound best when your digitized source has the highest possible gain without clipping. Digital audio files that do not use their full amplitude range produce low-quality streaming clips. If the amplitude range is too low, use your sound editor to adjust the range and increase the amplitude.
|
Tip: Most sound editors have a Normalize function that maximizes levels automatically. Because some systems have trouble with files normalized to 100 percent, normalize to 95 percent of maximum, or to -0.5dB. |
Normalization maximizes the volume level of the audio file's loudest sections. Consequently, quiet sections may not encode as well. Dynamics compression evens out input levels by attenuating (turning down) the input when it rises above a specified threshold. Check your audio software for a Compression or Dynamics feature. You can control attenuation by specifying a compression ratio. This turns down the loudest sections, and you can readjust input levels accordingly.
|
Tip: For multipurpose dynamics compression, set the threshold to -10dB, the ratio to 4:1, and the attack and release times to 100ms. Adjust the input level to get approximately 3dB of compression and an output level of about 0dB. |
Equalization (EQ) changes the tone of the incoming signal by "boosting" (turning up) or "cutting" (turning down) certain frequencies. Using EQ, you can emphasize certain frequencies and cut others that contain noise or unwanted sound. EQ can compensate for RealAudio codecs that do not have flat frequency responses (that is, codecs for which certain frequencies are not as loud after encoding). You can therefore use EQ to make a RealAudio clip sound as close as possible to the source recording.
|
Tip: For voice-only content, you can make the file more intelligible by cutting frequencies below 100 Hz and carefully boosting frequencies in the 1 to 4 kHz range. |
|
©2004 RealNetworks, Inc.
For more information, visit RealNetworks Click here if the Table of Contents frame is not visible at the left side of your screen. |