For decades, the MP3 marked a turning point in audio distribution by flooding devices with entire albums that took up a fraction of the space of WAV files. Today, that storage limitation is meaningless, since even the most basic cell phones typically exceed 64 GB, and bandwidth continues to grow unabated. Furthermore, encoding techniques have advanced to the point where they capture nuances that the MP3 format sacrifices—nuances that are indeed perceptible. In contrast, new audio formats take advantage of more robust processing architectures and allow for more granular compression settings, adapting to each user’s needs—whether streaming, archiving, or real-time transmission—without losing precision. With these factors in play, sticking with the MP3 tradition becomes a purely nostalgic choice. So, let’s take a closer look at each of the current audio codecs.
FLAC: bit-for-bit fidelity and comprehensive control
The Free Lossless Audio Codec stands out for fully preserving the original data, providing byte-for-byte identical copies of the master file. Under the hood, FLAC divides the signal into blocks of up to 16,384 samples, each processed using adaptive linear prediction and Rice coding to eliminate redundancies.
The compression levels—from 0 to 8—adjust the size of each block; therefore, the higher the level, the more intelligent the parameter selection, but the greater the computational load during encoding. Decompression, based on integer operations, is very CPU-efficient, and latency remains below one millisecond even on modest hardware. Each block includes a CRC16 checksum and a 128-bit MD5 hash of the original audio, ensuring error detection and complete integrity.
As for the metadata system, FLAC uses Vorbis comments, but it also supports sample look-up tables and embedded album art in PNG or JPG format, which makes it easier to manage files in libraries. The container’s modularity ensures that, if a block cannot be read, the rest of the file remains intact—which is vital for large files and live streams.
By offering lossless music, it makes it the top choice for archiving music collections and for audiophiles.
AAC (M4A/M4B): Perceptual Refinement and Compatibility
Advanced Audio Coding is an evolution of the MP3 format, refining psychoacoustic analysis with more flexible windowing filters that reduce pre-echo and improve high-frequency resolution.
In practice, AAC uses 1024- and 960-sample windows with dynamic overlap, optimizing encoding based on the spectral complexity of the audio segment. Its joint stereo architecture takes advantage of massive channel modeling: it combines the common and difference signals to minimize redundancy as much as possible.
It supports sample rates from 8 kHz to 96 kHz and bit depths up to 24 bits, with variable or constant bit rates ranging from 16 kbps to 512 kbps. When embedded in MPEG-4 containers, it uses the M4A extension for unprotected music and the M4B extension for audiobooks, which include chapter markers and automatic resume.
Support for Low Complexity (LC), Main, and Scalable Sample Rate (SSR) profiles provides flexibility depending on the platform: LC is the most widely used due to its balance between quality and processing load. Decoding supports specialized Huffman tables and the Modified Discrete Cosine Transform (MDCT), which refines spectral reconstruction without incurring noticeable latency.
What's more, AAC offers nearly universal compatibility; almost all current devices, browsers, and operating systems support it out of the box. It is the most common format used by streaming services.
Opus and Vorbis
Opus combines two engines: SILK, which is based on linear prediction for speech, and CELT, which is optimized for music. For each frame—up to 120 ms—the system autonomously decides which technology to apply, or whether to combine both, ensuring consistent quality across bitrate ranges from 6 to 510 kbps.
The codec automatically adjusts the bit rate on a frame-by-frame basis and uses variable-length MDCT transforms (ranging from 2.5 ms to 120 ms), keeping algorithmic latency below 26 ms.
Thanks to its buffered delay control and FEC (Forward Error Correction), it is ideal for video calls and interactive streaming, where every millisecond counts. It is typically embedded in the Ogg container, although it is also supported in Matroska and WebM.
For its part, Vorbis uses lossy compression within the Ogg container and bases its scheme on fixed-size windows (1,024 or 2,048 samples) and MDCT coefficient vector encoding. It uses VBR with a quality scale ranging from -1 to 10, allowing the encoder to adjust the bitrate based on complexity.
At medium levels (level 5), it offers superior quality to MP3 at 192 kbps, and can scale up to 500 kbps for high-resolution audio at 44.1 kHz. Its lack of patent dependencies encourages open-source software projects and community-driven streaming platforms. Although it has taken a back seat to Opus and AAC in commercial settings, it remains alive and well in open-source applications, video games, and web services seeking to avoid royalties.
