Many different audio file formats exist for storing recorded audio data on a computer system. This post compares multiple file types and gives suggestions on which formats and bitrates one should use, especially when producing podcasts or other online audio.
We also added two listening examples to compare MP3, AAC and Opus files at various bitrates.
If you just want to see some practical tips, skip the description and read the conclusion.
UPDATE 2017:
The original article was updated to include the new Opus audio codec, concrete bitrate suggestions, audio examples at various bitrates and other recent developments!
Types of audio formats
There are three main types of audio file formats:-
Uncompressed audio formats:
Uncompressed audio formats store the audio information as it is recorded. This results in big files, but no information is lost, therefore they are suitable for archiving original recordings. The most common uncompressed audio format is PCM, which is usually stored in a WAV or AIFF file. -
Lossless compression:
Lossless audio compression formats need less space then uncompressed formats, without any loss in quality. They work similar to ZIP, but the compression algorithms are specifically designed for audio data. Some example formats are FLAC (Free Lossless Audio Codec) or ALAC (Apple Lossless Audio Codec), for a comparison of various codecs see Lossless comparison. -
Lossy compression:
Lossy compression formats significantly reduce the file size, by throwing away information imperceptible to humans. This gives very small files, but some information is lost and cannot be reconstructed. The best-known example is MP3, others are AAC, Opus, Ogg Vorbis, WMA, ...
The Bitrate defines, how many bits (storage) are used to encode a certain amount of audio - for example, 128 kbps will use about 128 kilobits for each second of audio that is encoded. In Constant Bitrate Encoding (CBR), the bitrate is kept constant across the entire file, whereas Variable Bitrate Encoding (VBR) tries to maintain a constant quality by choosing the optimal bitrate to represent each audio frame (for example: a higher bitrate for complex audio, much lower bitrate for silence).
Joint Stereo methods further reduce the file size of Stereo Signals, so that the quality of a 128kbps Stereo MP3 file is usually much higher than a 64kbps Mono MP3 file!
A great source of information about audio codecs is the hydrogenaudio.org wiki and forum.
Relevant audio formats for podcasts
When distributing a podcast or other audio over the internet, you want to have the smallest possible filesize, the best possible
quality and everyone should be able to play it (on all operating systems, on mobile phones, portable audio players, car audio players etc.).
Because of the much smaller filesize, lossy formats are the only real option.
Additionally one may archive the produced podcast in a lossless compressed audio file.
The characteristics of a few important formats are listed below.
We also suggest bitrates for spoken word audio. Note that the bitrate depends on your audio material and should be higher if your audio contains a lot of music!
- the most widespread and the de facto standard of online audio file formats
- bad quality at low bitrates (<= 80kbps)
- podcasts use constant bitrate encoding (CBR), because of seeking problems with VBR
- a license is required to "distribute and/or sell decoders and/or encoders" (see e.g. MP3 licensing) - however, MP3 should be patent free on 30 December 2017
- suggested bitrate for podcasts: 112kbps for stereo, 80kbps for mono audio
- the industry standard and the official successor to MP3 (also called MP4 audio, most common file extension is M4A)
- widespread in the Apple and broadcast community (iTunes, iPod, iPhones, set-top boxes etc.), also supported on Android and most desktop audio players
- better quality and lower bitrates compared to MP3
- different types of AAC exist (LC AAC, HE AAC, HE AAC v2, etc.)
- HE AAC uses spectral band replication (high frequencies are removed and calculated from lower frequencies, see SBR), HE AAC v2 adds a parametric stereo method (stereo audio is created out of a mono signal, see PS) - don't use HE AAC for higher bitrates!
- heavily patented: a patent license is required for all manufacturers or developers of AAC codecs
- suggested bitrate for podcasts: 64/80kbps for stereo, 40/48kbps for mono audio
- the latest and most advanced open source/patent free audio codec
- supported by all major web browsers (WebRTC), desktop audio players and on Android 5+, no native iOS support
- separate speech (SILK) and music (CELT) codec, which makes it especially interesting for podcasters
- performs well also at very low bitrates
- suggested bitrate for podcasts: 56/64kbps for stereo, 32/40kbps for mono audio
- ideal format for archiving
- lossless compression, no information is lost
- flac files are typically reduced to 40-60% of their original size
- very fast encoding and decoding
- open source and patent free
All formats listed here support common metadata tags, including Chapter Marks / Enhanced Podcasts. However, the support for chapters depends on the used audio player, for details please see Which Media Players Support Chapter Marks.
A detailed discussion about audio formats for podcasts in German can be found in the podcast episode LS018 Audioformate für Podcasts.
Audio Examples at various Bitrates
We created one mono and one stereo audio example file encoded in MP3 (constant bitrate), in AAC and in Opus to compare the audio quality at low bitrates.
You will need a recent Firefox or Chrome browser to listen to all example files.
Because it's very hard to hear any artifacts at higher bitrates, we did not include high bitrate audio files.
More details about these audio examples (encoder parameters, subjective impressions, etc.) are available in the comments of the following G+ post:
Audio Examples: MP3 vs. AAC/M4A vs. Opus.
1. Stereo Example:
This stereo file from Bits of Berlin starts with a music intro, then contains female and male speakers.
Click on the links to download or play the audio files.
Uncompressed WAV audio source:
WAV audio
2. Mono Example:
This mono file from Bits of Berlin contains female and male speakers.
Click on the links to download or play the audio files.
Uncompressed WAV audio source:
WAV audio
For a more detailed and scientific comparison of various codecs at different bitrates see List of Codec Listening Tests.
Conclusion
If you want to reach every old device out there and maintain only one file format, then use MP3 (CBR - constant bitrate encoding). It is not the most advanced codec, but everything supports it.
Additionally it may be advantageous to offer your audio in AAC and Opus.
AAC is the standard on all Apple and broadcast devices, used for Apples
Enhanced Podcasts
and shows good performance especially at low bitrates (HE AAC).
Opus is the most advanced and latest open source audio codec and is supported by all major browsers, which makes it
especially interesting for HTML5 podcast web players (high quality at lowest possible bitrate).
Finally, if you also want to archive an original version of your audio, use FLAC, ALAC or just PCM (WAV, AIFF).