Published: 2026-06-05 · Reading time: 11 min read · Author: Nikolay Sapunov, CEO at Fora Soft

Why this matters

If you build an OTT service, a music app, a post-production tool, or any product that stores original audio before it gets compressed for delivery, you will have to choose a format for the high-quality master — and the lossless family is where that choice lives. This article is written for a product manager, founder, or operations lead with no audio background: by the end you will understand what "lossless" actually means, how FLAC, ALAC, and WavPack differ, why they all come out roughly the same size, and exactly when your product needs lossless instead of the cheaper lossy codecs. Every spec number here traces back to its controlling document — for FLAC, that is now a formal IETF standard — not a secondhand blog summary.

What "lossless" really means

A codec is an agreed method for compressing sound on one end of a link and reconstructing it on the other. Most codecs a video product uses — AAC, Opus, MP3 — are lossy: they permanently discard parts of the signal a listener is unlikely to notice, and the decoded audio is similar to the original but not identical. A lossless codec makes a different promise. It compresses the audio, and when you decode it you get back the original data exactly, bit for bit, with nothing missing. The trade is right there in the name: you lose nothing, but you save less space.

The numbers make the trade concrete. A music codec like Opus, covered in the Opus article, can carry a stereo song at 128 kbit/s. The same song stored losslessly typically runs around 700–900 kbit/s — five to seven times larger. So lossless is never the cheap option. It is the faithful option, and you pay for fidelity in bytes.

How does a lossless codec save any space at all if it throws nothing away? It exploits predictability. Real audio is not random: the next sample is usually close to the last few. A lossless encoder builds a small mathematical model that predicts each upcoming sample from the ones before it, then stores only the tiny difference between the prediction and the truth — a number that is almost always much smaller than the original sample, and therefore needs fewer bits. Decode the differences, run the same model forward, and the original returns exactly. This idea — predict, store the error, reverse it — is the engine inside all three formats below. The same family of ideas appears in how audio compression works, and the raw uncompressed starting point is the PCM described in what is digital audio.

Side-by-side comparison of lossy and lossless audio compression: a lossy codec like Opus shrinks a song to about 128 kbit/s by permanently discarding inaudible detail so the output differs from the input, while a lossless codec like FLAC shrinks the same song to about 800 kbit/s and returns a decoded file that is bit-for-bit identical to the original. An arrow shows lossless is five to seven times larger but exact Figure 1. The trade in one picture: lossy is small but approximate; lossless is larger but bit-exact.

FLAC: the open standard everyone can use

FLAC — the Free Lossless Audio Codec — is the most widely deployed lossless format and the one with the most open governance. For most of its life it was defined only by documentation on the Xiph.org website, but in December 2024 it became a formal internet standard, IETF RFC 9639, on the Standards Track (RFC 9639, December 2024). That matters for a product team: a codec defined by a stable, citable standard is a codec you can build on without fear of it shifting under you.

FLAC's reach is wide. It codes from 1 to 8 channels, sample rates from 1 Hz up to 1,048,575 Hz, and bit depths from 4 to 32 bits per sample — far more than any real recording needs, which is the point: it will never be the format that can't hold your audio (RFC 9639, §1). Internally it works in four steps. First it cuts the stream into short blocks of up to 65,535 samples. Then, for stereo, it removes redundancy between the two channels by storing a mid-and-side representation instead of plain left-and-right — because the two channels of a stereo recording are usually very similar, the difference between them is small and cheap to store (RFC 9639, §4.2). Next it runs the predictor described above, using either one of five fixed formulas or a custom linear-prediction model with up to 32 coefficients. Finally it packs the leftover prediction errors with Rice coding, an entropy method tuned for streams of small numbers with the occasional outlier (RFC 9639, §4.4).

Two practical features make FLAC the safe default. It carries an MD5 checksum of the original audio in its header, so a decoder can prove the file is intact — useful for an archive you will trust for a decade (RFC 9639, §8.2). And it defines a streamable subset: each frame carries enough information that a player can start decoding from the middle of a stream without reading the file header, which is exactly what an HTTP streaming service needs. That is why FLAC is the codec behind Tidal (which dropped the proprietary MQA format for open FLAC), Qobuz, and Amazon Music, all delivering up to 24-bit / 192 kHz today.

ALAC: Apple's lossless, for Apple's world

ALAC — the Apple Lossless Audio Codec — is Apple's own answer to FLAC, and for years the reason it existed was simple: Apple's players didn't decode FLAC, so the Apple ecosystem needed its own lossless format. ALAC compresses to a bit-for-bit identical copy of the original, just like FLAC, and supports 16-, 20-, 24-, and 32-bit audio in 1 to 8 channels (Apple ALAC project). Apple released the encoder and decoder as open source under the Apache 2.0 license in October 2011, so ALAC is no longer locked to Apple's platforms — but in practice it still lives almost entirely inside them.

The defining technical fact for a product team is the container. ALAC normally lives inside an MP4 / .m4a file, the same ISO base-media container that holds AAC, using a small block of setup data Apple calls a "magic cookie" (Apple ALAC project). Containers are covered in audio in containers. This is why ALAC is the codec behind Apple Music Lossless and Hi-Res Lossless, which stream up to 24-bit / 192 kHz to Apple devices at no extra subscription cost. If your product targets the Apple ecosystem or has to match Apple Music's lossless tier, ALAC is the format your masters and your delivery path will speak.

WavPack: the flexible one with a hybrid trick

WavPack, created by David Bryant and shipping since 1998, is the least famous of the three and the most versatile. In its normal mode it is an ordinary lossless codec like the other two. Its signature feature is hybrid mode, which no mainstream rival offers: instead of one file, WavPack can produce two — a small, self-contained lossy file (.wv) that plays fine on its own, plus a correction file (.wvc) that, added back, restores the full lossless original (WavPack documentation, 2024).

That split is genuinely useful in a production pipeline. You can ship or stream the lightweight .wv for everyday playback and keep the .wvc correction file in the archive, so you hold a perfect master without storing a separate full-size lossless copy. WavPack also handles more exotic inputs than the others — including 32-bit floating-point audio, the format modern editing tools capture in — which is why it shows up in mastering and archival workflows more than in consumer streaming.

They're all lossless — so how do you choose?

Here is the fact that surprises people: because all three are mathematically lossless, they all sound exactly the same — identical to the original and identical to each other. There is no "better-sounding" lossless codec, only smaller or larger files and faster or slower decoding. Independent testing on the audio-engineering community Hydrogenaudio puts FLAC and WavPack within a fraction of a percent of each other on file size (about 61% of the original for both), with ALAC slightly larger and slower to decode than either.

Format Status / license Native container 2026 home Special trait
FLAC IETF RFC 9639, open .flac, also fMP4 Tidal, Qobuz, Amazon Music Built-in MD5 integrity check
ALAC Apache 2.0, open since 2011 MP4 / .m4a Apple Music Lossless Native to the Apple ecosystem
WavPack Open source (BSD-style) .wv (+ .wvc) Mastering, archival Hybrid lossy + correction file

Because the audio is the same, the decision is driven by where the file has to play, not by quality. Targeting Apple devices or Apple Music parity? Use ALAC. Building an open, cross-platform music tier or a hi-res streaming catalog? Use FLAC — it is the de-facto industry standard and the only one defined by a public RFC. Need 32-bit float support or the hybrid lossy-plus-correction workflow for an archive? Use WavPack.

Common mistake: shipping lossless to listeners who can't hear the difference

The most expensive lossless error is using it for ordinary delivery. A lossless stream is five to seven times the size of a transparent lossy stream, and on phone speakers, laptop speakers, or typical earbuds, listeners cannot tell the two apart in a blind test. Paying for lossless bandwidth to every viewer of a video product is a cost with no perceptible benefit. The right pattern is to keep your masters lossless and deliver lossy — encode the high-quality lossless master once, then transcode to AAC or Opus for streaming. Reserve lossless delivery for a deliberate, opt-in audiophile or music tier, where the user has chosen it and often has the hardware to benefit.

Where Fora Soft fits in

In the video products we build — OTT and internet-TV platforms, e-learning systems, and media tools — the lossless decision usually shows up at the storage layer rather than the delivery layer. A streaming service keeps a lossless or near-lossless mezzanine so it can re-encode to new delivery codecs later without quality loss, while shipping AAC or Opus to viewers. A music-adjacent app that wants to match Apple Music's lossless tier has to speak ALAC inside MP4, while a cross-platform catalog leans on FLAC. Getting the master format right at the start saves a painful, lossy-by-accident re-encode of an entire library later.

What to read next

Call to action

References

  1. IETF RFC 9639 — "Free Lossless Audio Codec (FLAC)", M.Q.C. van Beurden & A. Weaver, Standards Track, December 2024, ISSN 2070-1721. The controlling specification for FLAC. Supports: 1–8 channels, 1–1,048,575 Hz, 4–32 bits; four-stage encode (blocking §4, interchannel decorrelation §4.2, prediction §4.3, Rice residual coding §4.4); streamable subset §7; MD5 checksum in STREAMINFO §8.2. https://www.rfc-editor.org/rfc/rfc9639.html
  2. Apple Lossless Audio Codec project (macOSForge / GitHub), Apple Inc., source released under Apache License 2.0, October 2011. Supports: bit-for-bit identical decode, MP4/M4A "magic cookie" carriage, 16/20/24/32-bit, 1–8 channels. https://macosforge.github.io/alac/
  3. WavPack 5 Library Documentation, David Bryant, February 2024, and WavPack 4 & 5 File Format, April 2020. Supports: hybrid mode (.wv lossy + .wvc correction), 8/16/24/32-bit integer and 32-bit float PCM. https://www.wavpack.com/WavPack5LibraryDoc.pdf
  4. Hydrogenaudio Knowledgebase — Lossless comparison. Independent community benchmark: FLAC ≈ 61.05% and WavPack ≈ 60.83% of original size; ALAC larger and slower to decode than both. Used as a tier-6 community source for the relative-size claim; absolute spec facts above defer to the standards documents. https://wiki.hydrogenaudio.org/index.php?title=Lossless_comparison
  5. Xiph.org FLAC — Comparison, the FLAC project's own benchmark page. Supports FLAC's fast-decode and wide-hardware-support claims. https://xiph.org/flac/comparison.html
  6. What Hi-Fi? — Hi-res music streaming services compared (2026). Supports 2026 deployment facts: Apple Music Lossless/Hi-Res uses ALAC up to 24-bit/192 kHz; Tidal, Qobuz, and Amazon Music use FLAC up to 24-bit/192 kHz; Tidal replaced MQA with FLAC. https://www.whathifi.com/advice/hi-res-music-streaming-services-compared
  7. ISO/IEC 14496-12 — ISO base media file format. The container standard underlying MP4/M4A that ALAC uses for carriage. https://www.iso.org/standard/83102.html
  8. Apple Lossless Audio Codec — Hydrogenaudio Knowledgebase. Supports ALAC channel-layout and bit-depth detail (1–8 channels, up to 32-bit). https://wiki.hydrogenaudio.org/index.php?title=ALAC