LESSON 19 · DIGITAL AUDIO
Does lossless sound better than MP3?
FLAC can preserve every audio sample. MP3 and AAC make smaller files by leaving information out. But what you can hear depends on the encoding, the music, and how you compare them, not merely the number printed beside a file.
Try our short listening test with two real songs. Play the original, switch to MP3, and see what you notice. No special knowledge needed.
Listen and compare MP3 vs lossless → Try a blind A/B/X listening game →- Bitrate
- The number of data bits stored or transmitted each second, across the encoded stream.
- Lossless
- Compression that restores the same original digital audio samples when decoded.
- Lossy
- Compression that makes irreversible changes to reduce how much data is needed.
What do these exact letters and numbers say? Decode the specification ↓
DECODE THE PRINTED SPECIFICATION
MP3 · 320 kb/s · stereo · CBR
320 kb/s means about 320,000 encoded bits in one second, total.
MP3 names the compression format. Stereo indicates two-channel playback. CBR means constant bitrate: the stream aims to use the same data rate throughout, rather than freely varying its average. 320 kb/s describes the combined encoded stream, not 320 kb/s for each channel.
Here's what 320 kb/s means for storage, not hearing: a file uses roughly 320,000 bits per second. Eight bits make one byte, so that's about 40,000 bytes each second, 2.40 MB per minute, or 9.60 MB for four minutes, before small file overhead. More bits can preserve more information, but the number alone cannot tell you what differences you'll hear.
What this number doesn't tell you: whether the recording was mastered differently, what the encoder removed, or whether any change is audible to you. Bitrate tells you the amount of data used each second, not a guaranteed sound-quality score.
320 kb/s shown on a 0–400 kb/s ruler
One second of encoded data, split into eight equal conceptual portions
The blocks illustrate the arithmetic, not real MP3 frames or a measured audio waveform. kB and MB are decimal units. Tiny headers and tags can change file size.
COMPARE WHAT IS STORED
How much data does a four-minute track use?
Switch between four common illustrative encoding specifications. The chart keeps the same 0–1,500 kb/s scale. The size numbers come from bitrate × time, not from encoding or listening to a real song.
Lossy compressed
MP3 · 320 kb/s CBR
A high-bitrate MP3 example. It stores about 22.7% as many stream bits as the uncompressed PCM example. Some of the source information is intentionally omitted.
Storage bitrate · one fixed 0–1,500 kb/s ruler for every example
About 22.7% of the uncompressed PCM example’s bit rate. This ratio describes stored data, not the percentage of music retained.
Show the four-minute calculation
320,000 bits/s × 240 s ÷ 8 = 9,600,000 bytes. Figures use decimal MB (1 MB = 1,000,000 bytes) and exclude container headers, tags and other file overhead. AAC example uses an assumed average rate; real variable-rate files may differ.
Lossy compressed
MP3 · 128 kb/s CBR
A smaller MP3 example that gives the encoder less room to describe challenging musical events. Audible differences remain track-, encoder- and listener-dependent.
Storage bitrate · one fixed 0–1,500 kb/s ruler for every example
About 9.1% of the uncompressed PCM example’s bit rate. This ratio describes stored data, not the percentage of music retained.
Show the four-minute calculation
128,000 bits/s × 240 s ÷ 8 = 3,840,000 bytes. Figures use decimal MB (1 MB = 1,000,000 bytes) and exclude container headers, tags and other file overhead. AAC example uses an assumed average rate; real variable-rate files may differ.
Lossy compressed
AAC · example 256 kb/s
A separate lossy codec, shown at a hypothetical fixed-average bitrate for simple arithmetic. The same numeric bitrate does not imply the same quality across codecs.
Storage bitrate · one fixed 0–1,500 kb/s ruler for every example
About 18.1% of the uncompressed PCM example’s bit rate. This ratio describes stored data, not the percentage of music retained.
Show the four-minute calculation
256,000 bits/s × 240 s ÷ 8 = 7,680,000 bytes. Figures use decimal MB (1 MB = 1,000,000 bytes) and exclude container headers, tags and other file overhead. AAC example uses an assumed average rate; real variable-rate files may differ.
Uncompressed
PCM · 16-bit / 44.1 kHz stereo
The raw CD-style PCM reference. This rate follows from 16 bits per sample × 44,100 samples per second × two channels. It is not itself a claim about mastering quality.
Storage bitrate · one fixed 0–1,500 kb/s ruler for every example
About 100% of the uncompressed PCM example’s bit rate. This ratio describes stored data, not the percentage of music retained.
Show the four-minute calculation
16 bits × 44,100 samples/s × 2 channels × 240 s ÷ 8 = 42,336,000 bytes. Figures use decimal MB (1 MB = 1,000,000 bytes) and exclude container headers, tags and other file overhead. AAC example uses an assumed average rate; real variable-rate files may differ.
Where is FLAC on that chart?
FLAC deliberately has no fixed bar here. It is lossless compressed PCM, and its average bitrate changes with the music and encoding settings. A simple passage may compress very differently from a dense or noisy one. Assigning FLAC one universal bitrate would make a neat graph and teach the wrong lesson.
THREE WAYS TO STORE THE SAME SOURCE
PCM, FLAC and MP3 are doing different jobs.
Imagine that your original digital recording is a long page of numbers. Uncompressed PCM stores those numbers directly. FLAC stores a smarter, smaller description that can reproduce every one of those original audio numbers when decoded. MP3 or AAC reduce the data by making decisions about sound information that is less likely to be noticed.
Lossless doesn't mean “better mastered.” A FLAC of a badly clipped master is still a perfectly faithful FLAC of that clipped master. Likewise, a carefully mastered MP3 can be more enjoyable than a harshly mastered FLAC of a different release.
WAV and AIFF usually carry uncompressed PCM, while .m4a is a container that might hold AAC or lossless ALAC, depending on the file. Check the actual codec, not just the filename extension. A 24-bit FLAC and a 16-bit FLAC are both lossless relative to the PCM audio they were made from, even though their underlying samples differ in precision.
Converting MP3 to FLAC does not reverse the loss. The new FLAC perfectly preserves the already-decoded MP3 result, including its existing artifacts. It does not restore the missing original information.
WHAT LOSSY CAN CHANGE
What should you actually listen for?
Lossy encoders try to remove or approximate information that the listener is unlikely to notice, using models of hearing and masking. At constrained bitrates or on difficult recordings, the approximations can become audible. They are called compression artifacts.
Pre-echo and smeared attacks
A very sharp drum or percussion hit can develop a faint “shh” just before the hit, or lose some crispness. Not every softened drum was damaged by a codec; the original recording might simply sound that way.
Swishy or warbling textures
Cymbal tails, ambience or sustained tones can sometimes sound watery, gritty or unstable as an encoder struggles with complex high-frequency material.
Less stable spatial cues
At sufficiently constrained settings, some stereo ambience or positioning detail may be simplified. That does not mean every lossy file has a narrow soundstage.
Upper-frequency changes
Some encoder settings apply low-pass filtering to spend bits where they may be more audible. A visible frequency cutoff alone does not reliably prove audible quality or reveal the original encoding history.
As bitrate and encoder quality improve, artifacts generally become harder to notice, but there is no magical bitrate where every recording becomes audibly identical for every listener. The specific encoder, version, settings, source and listener all matter.
This connects to the glossary's grainy, airy and separation descriptions. A subjective word is a clue to investigate, not proof of codec damage.
Can you reliably hear lossless versus a good MP3?
Sometimes, depending on the track, encoder, bitrate, equipment and listener. In other situations, a good lossy copy may be perceptually transparent: you cannot reliably distinguish it from the matched reference under controlled conditions. The file being smaller does not establish an audible problem.
For a fair comparison, start with the exact same master, encode the lossy copy directly from that lossless source, match playback levels, avoid visual cues, and compare blind. An ABX test asks whether an unknown “X” is A or B across repeated trials, so a claimed difference can be checked rather than merely described. The Listening Lab lets you compare real audio, and the ABX Training game lets you test your guesses without seeing which version is playing. These are listening examples, not proof that everyone will hear the same differences.
POPULAR CLAIM
“A 320 kb/s MP3 contains only 23% of the sound, so it must sound 77% worse.”
WHAT WE CAN ACTUALLY SAY
The data rate is roughly 23% of CD-style PCM, but codec design is deliberately selective. Data-rate reduction is not a percentage of audible quality lost. Only suitable listening tests can establish whether differences are reliably audible.
What is CBR versus VBR? Optional detail
CBR aims for a constant encoded bitrate. VBR means variable bitrate, so the encoder can devote more or fewer bits to different parts of the recording. VBR files are often described by their overall average bitrate or by an encoder quality setting, not a guaranteed bitrate at every moment. Even two 256 kb/s files made by different encoders are not automatically equivalent in audible quality.
For uncompressed stereo PCM, 16 bits × 44,100 samples/s × 2 channels = 1,411,200 bits/s. For a 320 kb/s constant-rate stream, four minutes of audio contain about 9.60 MB of encoded data. See bit depth and sample rate for the two separate PCM terms.
Which format should you actually use?
- Archiving CDs or a music library? Prefer genuine lossless sources such as FLAC when storage permits. They preserve the decoded PCM and can be converted to other formats later without inherited lossy damage.
- Listening on limited storage or bandwidth? Well-encoded MP3, AAC and other modern lossy codecs can offer useful size savings. Choose a reasonable quality setting and verify compatibility rather than worshipping one number.
- Already have MP3 files? Keep the original files until you obtain genuinely lossless versions. Merely re-encoding MP3 to FLAC or a higher MP3 bitrate does not restore missing information, and repeated lossy re-encoding can compound damage.
- Judging an upgrade? Compare like-for-like masters under controlled levels, and distinguish compression artifacts from EQ, remastering, amplification, speaker placement and room effects.
Related lessons: 16-bit versus 24-bit, sample-rate limits and DAC performance numbers.
Sources and further reading
- Xiph.Org · What FLAC preserves and why it is lossless ↗
- Xiph.Org · How FLAC stores PCM and variable-rate frames ↗
- Audacity · Lossy versus lossless and perceptual transparency ↗
- iZotope · MP3/AAC encoding artifacts and peaks ↗
- foobar2000 · ABX comparator for double-blind listening ↗
Integrity note: Every size and bar on this page is a mathematical teaching example. No commercial audio excerpt, real encoder output, listening-test result or proprietary recording is presented as proof.