FLAC vs MP3

How FLAC compresses without losing data

FLAC (Free Lossless Audio Codec) was developed by Josh Coalson in 2001, now maintained by Xiph.Org. The compression algorithm:

1. Block the audio into frames (typically 4096 samples)
2. For each frame, fit a linear predictor (predicts each sample from previous samples in the same channel)
3. Compute the residual (actual sample minus predicted)
4. Encode the residual with Rice coding (efficient for small numbers, which residuals tend to be)
5. Add a CRC for integrity checking

Because step 3 is mathematically reversible (you can compute the predictor coefficients exactly from the residual), the original samples can be perfectly reconstructed. Hence "lossless".

The compression ratio depends on how predictable the audio is:

• Classical / acoustic music: highly predictable → FLAC compresses to ~40% of source
• Pop / rock with dense mixing: less predictable → FLAC compresses to ~50% of source
• White noise / very chaotic audio: incompressible → FLAC stays at ~95% of source (worst case)

For typical music, expect FLAC to be 50-60% of the original WAV size. CD audio (~10 MB/min) becomes ~5-6 MB/min as FLAC.

How MP3 throws away data you "can't hear"

MP3 (MPEG-1 Audio Layer III, ISO/IEC 11172-3) uses psychoacoustic perceptual coding:

1. Block the audio into frames (~26ms each)
2. Apply a hybrid filter bank (subband + MDCT) to convert to frequency domain
3. Calculate masking thresholds — frequencies that would be inaudible because louder frequencies nearby mask them, or because they're outside the audible range, or because of temporal masking from previous loud sounds
4. Quantize each frequency component with bits proportional to its perceptual importance (more bits for unmasked, audible components; fewer bits for masked)
5. Huffman-encode the quantized values

The trick: at 192-256 kbps, the discarded information is statistically inaudible to most listeners on most equipment in most situations. At 320 kbps (MP3 maximum), even trained audio engineers struggle to reliably distinguish from source in blind tests.

But here's the catch: the discarded information is gone forever. You can decode MP3 → re-encode to MP3 again at 320 kbps → still missing the original information. This is "generation loss" in lossy formats.

Bitrate equivalence in real listening

For typical music on consumer headphones:

On audiophile gear (high-end headphones, dedicated DAC, treated room): trained listeners can sometimes detect 256 kbps MP3 vs FLAC, especially on complex orchestral pieces or dense electronic music with subtle high-frequency detail.

On phone earbuds, Bluetooth speakers, car stereos, laptop speakers: the difference is essentially imperceptible at 192+ kbps.

File size: real comparison for a 4-minute song

Source: 16-bit / 44.1 kHz CD audio = 44.1 MB raw (WAV)

FLAC is roughly 3× larger than 320 kbps MP3 and 4× larger than 192 kbps MP3 for the same content.

When FLAC actually matters: the honest list

Despite all the above, there are real situations where FLAC's bit-perfect quality matters:

1. Editing workflow — Multiple generations of edits (cut, re-mix, normalize, etc.) compound lossy degradation. FLAC stays pristine forever.
2. Re-encoding to other lossy formats — If you need to convert your library to a different lossy format in the future (say, MP3 → Opus for streaming), going from FLAC source produces better quality than going from MP3.
3. Audiophile playback on capable systems — Studio monitors + dedicated headphone amp + treated room: subtle differences become detectable.
4. Hi-Res audio (24-bit / 96-192 kHz) — Some recordings released in higher-than-CD resolution. MP3 cannot represent this.
5. Surround sound music — FLAC handles 5.1/7.1; MP3 is essentially stereo.
6. Long-term archival — In 50 years, audio compression will have evolved many generations. FLAC will still play (open spec, well-documented). The MP3 you ripped in 2007 might be readable but won't benefit from any improvements.

When MP3 is genuinely the right choice

1. Mobile listening — Storage and battery life matter. MP3 wins.
2. Streaming non-Hi-Res tiers — Spotify, Apple Music standard, YouTube Music: lossy delivery anyway.
3. Voice content (podcasts, audiobooks) — Voice compresses well at low bitrate. FLAC is overkill.
4. Universal compatibility — Old hardware, mixed-audience sharing.
5. Bandwidth-constrained scenarios — Streaming over cellular, slow internet.

Converting between FLAC and MP3

Convert FLAC to MP3 extracts the lossless audio and re-encodes to MP3 at a chosen quality. You lose quality (lossless → lossy is a one-way street). Quality 192-256 kbps gives "audibly identical" results for most listeners.

Convert MP3 to FLAC packages the existing MP3 audio in a FLAC container without losing further quality, but doesn't restore the data the original MP3 encoding discarded. The resulting FLAC is 3-5× larger than the source MP3 but contains the same audible information. Useful only when an upstream pipeline requires FLAC input; otherwise wasteful.

For best quality, always work from the original lossless source (CD, FLAC, WAV) when producing any compressed format. Never re-encode lossy to lossy if you can avoid it.