AAC (Advanced Audio Coding) is a lossy perceptual audio codec standardized by MPEG as the successor to MP3. Among audio formats and codecs, it is one of the most widely deployed, serving as the default audio format of MP4 video, most streaming services, mobile devices and several digital broadcast systems.
AAC was first specified in MPEG-2 Part 7 in 1997 and extended in MPEG-4 Part 3, which added a family of profiles for very low bitrates, low delay and unified speech and audio coding. At comparable bitrates, it generally delivers better quality than MP3, particularly below about 128 kbit/s for stereo.
How AAC compresses audio
AAC is a transform codec. It converts blocks of audio samples into frequency coefficients with the modified discrete cosine transform (MDCT), quantizes those coefficients according to a psychoacoustic model and stores them with entropy coding. Information judged inaudible is discarded, which is why the codec is lossy.
The encoder normally uses long windows of 2048 samples that yield 1024 coefficients per frame, giving fine frequency resolution for stable tones. Around sharp transients, it switches to eight short windows of 256 samples, which limits pre-echo, the smearing of noise before a sudden attack.
Several coding tools improve efficiency beyond MP3. Temporal noise shaping shapes quantization noise in time within a frame, joint stereo coding exploits similarities between channels, and a more flexible filter bank handles both tonal and percussive material better than the hybrid filter bank of MP3.
AAC profiles and their purposes
AAC is a family of profiles rather than a single algorithm. Each profile adds or removes tools to suit a particular bitrate range or latency requirement.
| Profile | Key tools | Typical stereo bitrate | Main use |
|---|---|---|---|
| AAC-LC | Core MDCT coding | About 96 to 256 kbit/s | Music, video soundtracks, downloads |
| HE-AAC | Adds spectral band replication (SBR) | About 48 to 80 kbit/s | Streaming radio and adaptive streaming |
| HE-AAC v2 | Adds parametric stereo | About 24 to 48 kbit/s | Very low bitrate and mobile broadcast |
| AAC-LD and AAC-ELD | Short frames, low delay | About 48 to 128 kbit/s | Real-time communication |
| xHE-AAC | Unified speech and audio coding | From about 12 kbit/s upward | Adaptive streaming of speech and music |
Spectral band replication transmits only the lower part of the spectrum in detail and reconstructs the high frequencies from it using a small amount of side information. Parametric stereo sends a mono signal plus parameters that recreate the stereo image. Both tools trade fidelity for much lower bitrates.
How AAC compares with other audio codecs
AAC competes with MP3, Opus and Ogg Vorbis among lossy codecs and coexists with lossless formats such as FLAC and ALAC for archiving and high-fidelity distribution.
| Format | Type | Strength | Limitation |
|---|---|---|---|
| AAC | Lossy | Universal hardware support, good efficiency | Patent licensing for some uses |
| MP3 | Lossy | Universal compatibility | Less efficient at low bitrates |
| Opus | Lossy | Excellent at low bitrates and low latency | Less common in broadcast and older devices |
| Ogg Vorbis | Lossy | Open and royalty-free | Limited hardware support |
| FLAC and ALAC | Lossless | Bit-exact reproduction | Files several times larger |
| WAV | Uncompressed PCM container | Simple and universal in production | Largest files |
In practice, AAC remains the safest choice for compatibility in video and consumer playback. Opus often wins in interactive communication and web real-time applications, where low latency and open licensing matter more.
Containers and practical encoding choices
AAC data is carried in several container and transport formats. Choosing the right wrapper matters as much as choosing the bitrate, because players and broadcast chains expect specific formats.
| Situation | Container or transport | Recommended setting |
|---|---|---|
| Music download or library | MP4 audio (.m4a) | AAC-LC at 192 to 256 kbit/s stereo |
| Video file for the web | MP4 or MOV | AAC-LC at 128 to 192 kbit/s stereo |
| Adaptive streaming | fMP4 or MPEG-TS segments | HE-AAC for low rungs, AAC-LC for high rungs |
| Broadcast and contribution | MPEG-TS with ADTS or LATM | AAC-LC or HE-AAC as specified by the operator |
| Surround soundtrack | MP4 or MPEG-TS | AAC-LC 5.1 at about 320 to 384 kbit/s |
Encoder quality varies considerably. Two encoders at the same bitrate can produce audibly different results, so the encoder implementation and its mode, constant or variable bitrate, matter as much as the nominal number.
Limitations and common mistakes
AAC is efficient, but it remains lossy and generationally fragile. Each decode and re-encode cycle adds new artifacts, so AAC should be used for delivery rather than as an editing or archive format.
- Transcoding from another lossy format: converting MP3 to AAC compounds artifacts instead of improving quality.
- Using HE-AAC at high bitrates: above roughly 96 kbit/s stereo, AAC-LC usually sounds better.
- Ignoring encoder delay: AAC adds priming samples at the start, so gapless playback needs correct metadata.
- Assuming decoder support for every profile: older devices may play only AAC-LC, or ignore SBR and output reduced bandwidth.
- Clipping on decode: masters at 0 dBFS can exceed full scale after encoding, so a true-peak ceiling of about -1 dBTP is recommended.
In everyday language
AAC is a way of making audio files much smaller while keeping them sounding good. It removes parts of the sound that people are unlikely to notice. A typical AAC song is therefore only a fraction of the size of the original studio recording.
It is the audio inside most online videos, streaming apps and phone recordings. Compared with the older MP3 format, AAC usually sounds better at the same file size. That is why so many devices and services use it.
Seen in the real world
A video platform prepares a concert film for adaptive streaming. The engineering team creates several audio renditions from the uncompressed master: HE-AAC at 64 kbit/s for viewers on weak mobile connections and AAC-LC at 192 kbit/s for viewers on fast home broadband. A 5.1 version for smart televisions is encoded with AAC-LC at 384 kbit/s.
Before encoding, the master is limited to -1 dBTP so that decoded peaks do not clip. The player switches audio renditions together with video quality. Viewers on weak connections hear clear sound without buffering, while viewers on fast links receive audio that is difficult to distinguish from the master in normal listening.