What a bitrate actually is
The bitrate states how much data is stored per second of audio. It is measured in kilobits per second (kbps). An MP3 at 128 kbps uses 128,000 bits per second. That is 16 kilobytes per second, or roughly 0.96 MB per minute. At 320 kbps, the rate is 40 kB/s or 2.4 MB per minute – 2.5 times as much.
For comparison: on two channels, an uncompressed audio CD delivers 44,100 samples per second at 16-bit resolution. This works out to 44,100 × 16 × 2 = 1,411,200 bits per second, i.e. 1,411 kbps. Therefore, a 128 kbps MP3 keeps only about 9% of the original data. A 320 kbps MP3 keeps about 23%. Psychoacoustics is the reason this can sound acceptable at all – more on that in a moment.
One thing to keep straight: what the bitrate measures is data quantity, not sound quality directly. How good a file sounds at a given bitrate depends on the codec (MP3, AAC, Opus), on the encoder, and on the source material. That is why 128 kbps is not always 128 kbps.
File sizes compared
The math is simple: file size = bitrate × duration ÷ 8. For a typical 4-minute song (240 seconds):
| Format / bitrate | Per minute | 4-minute song | Album (60 min) |
|---|---|---|---|
| MP3 128 kbps | 0.96 MB | 3.8 MB | 58 MB |
| MP3 192 kbps | 1.44 MB | 5.8 MB | 86 MB |
| MP3 256 kbps | 1.92 MB | 7.7 MB | 115 MB |
| MP3 320 kbps | 2.40 MB | 9.6 MB | 144 MB |
| FLAC (lossless, ~850 kbps) | ~6.4 MB | ~25 MB | ~380 MB |
| WAV (CD quality, 1,411 kbps) | 10.6 MB | 42 MB | 635 MB |
FLAC compresses losslessly. Depending on the material, it typically lands at 50–70% of the WAV size – the table shows a typical average, not a constant. Quiet piano music compresses far better than densely mastered rock.
Going from 128 to 320 kbps costs 2.5 times the storage. Whether that is worth it, however, depends on whether the difference is audible at all – and that is where it gets interesting.
Lossy compression: what gets thrown away at 128 kbps
MP3, AAC and Opus are lossy codecs. A psychoacoustic model guides them as they deliberately discard parts of the signal that human hearing perceives poorly or not at all:
- Masking: A loud tone hides quieter tones at neighbouring frequencies (simultaneous masking) and shortly before/after it (temporal masking). These masked components are then coarsely quantised or removed entirely.
- Low-pass filtering: At low bitrates, encoders cut high frequencies. With the LAME encoder, the low-pass typically sits around 16 kHz at 128 kbps; at 320 kbps, it is only around 20 kHz. For context: most adults hear up to roughly 15–17 kHz, while teenagers hear up to nearly 20 kHz.
- Joint stereo: Mid and side signals are encoded instead of two independent channels – nearly lossless for most recordings, but a significant bit saver.
The lower the bitrate, the more aggressively these mechanisms kick in. Typical artefacts of 128 kbps MP3 include smeared cymbals and hi-hats, "swirly" treble, and pre-echo on sharp transients (castanets, applause, harpsichord). Pre-echo happens because MP3 works in blocks of 576 or 1,152 samples. Quantisation errors then get smeared across the whole block. This is a structural weakness of the 1993 format. More modern codecs handle it better with flexible block sizes.
MP3 vs AAC vs Opus: a bitrate is not just a bitrate
MP3 (MPEG-1 Audio Layer III) was standardised in 1993. Its last patents expired in 2017. AAC followed in 1997 as the designated successor; at the same bitrate, it is audibly more efficient. Opus (IETF RFC 6716, 2012) is the most modern of the three. It is royalty-free, scalable from 6 to 510 kbps, and regularly on top in blind tests at low and medium bitrates.
As a rough guide to which bitrates deliver similar perceived quality across the three codecs:
| Quality target | MP3 | AAC | Opus |
|---|---|---|---|
| Speech, clearly intelligible | 64–96 kbps | 48–64 kbps | 24–48 kbps |
| Music on the go, decent | 128–160 kbps | 96–128 kbps | 80–96 kbps |
| Transparent for most listeners | 190–245 kbps (V2–V0) | 160–192 kbps | 128–160 kbps |
| Typical encoder maximum | 320 kbps | 320 kbps (LC) | 510 kbps |
"Transparent" means: in a blind comparison, no longer reliably distinguishable from the lossless original. These figures come from public listening tests (including the Hydrogenaudio community) and are guidelines. Critical material, however, pushes the thresholds upwards. The practical consequence: even though the number is smaller, a 256 kbps AAC file is at least the equal of a 320 kbps MP3. That is exactly why Apple has sold 256 kbps AAC rather than 320 kbps MP3 since iTunes Plus (2007). If you need an M4A/AAC file for an old device that only understands MP3, you can convert it with M4A → MP3 – but pick a high target bitrate to keep the transcoding loss small.
Can you actually hear the difference?
The inconvenient truth from blinded ABX tests: most people cannot distinguish a well-encoded 192 kbps file from the lossless original. At 256 and 320 kbps, even trained listeners with studio headphones manage it only on selected problem samples – if at all.
Between 128 and 256 kbps, however, the difference is genuinely perceptible to many listeners – especially on cymbals, applause and dense treble. Three factors decide whether you will hear it:
- Playback chain: Over Bluetooth earbuds in street noise, the difference practically vanishes – not least because Bluetooth itself re-compresses lossily (SBC, AAC or aptX). In a quiet room over open studio headphones, it is far more likely to show.
- Material: A loudly mastered pop track is very forgiving. The classic encoder killers, however, are harpsichord, castanets, live applause and quiet classical passages.
- Hearing and training: You spot artefacts much earlier if you know what to listen for (pre-echo, swirly cymbals). Age matters too, since the upper hearing limit declines.
If you want to know for yourself: your own listening test can be run in ten minutes with an ABX tool (such as the foobar2000 ABX plugin). The result is often sobering – and it saves storage afterwards.
CBR vs VBR: constant or variable?
Besides the level of the bitrate, two modes of operation exist:
- CBR (constant bitrate): Every second gets the same number of bits – including silence. The file size is predictable, but the mode is inefficient: complex passages get too few bits, simple ones too many.
- VBR (variable bitrate): Bits are distributed by the encoder as needed. LAME VBR presets such as V0 (~245 kbps on average) or V2 (~190 kbps) deliver more consistent quality than CBR at the same average size.
For archiving and everyday use, VBR is the better choice. CBR still makes sense only where old hardware or streaming protocols demand a constant data rate.
What streaming services actually deliver
| Service | Codec | Maximum quality |
|---|---|---|
| Spotify | Ogg Vorbis / AAC, FLAC | 320 kbps ("Very High"); lossless FLAC since September 2025 |
| Apple Music | AAC / ALAC | 256 kbps AAC; lossless up to 24-bit/192 kHz (since 2021) |
| YouTube Music | AAC / Opus | 256 kbps AAC (Premium) |
| Amazon Music Unlimited | FLAC | "HD" 16-bit/44.1 kHz, "Ultra HD" up to 24-bit/192 kHz |
| Tidal | FLAC | Lossless up to 24-bit/192 kHz |
Notably, Apple is content with 256 kbps AAC as its standard tier – and nobody complains about the sound. This matches the test evidence: in everyday listening, an efficient codec at 256 kbps is indistinguishable from lossless.
The transcoding mistake: why 128 → 320 achieves nothing
A widespread misconception is "upconverting" a 128 kbps MP3 to 320 kbps to improve its quality. It does not work. Whatever the encoder threw away is gone for good. The new file is two and a half times as large, but at best it sounds identical. Usually it even sounds worse, because the second lossy encoding pass adds new artefacts (generation loss).
From this, two rules follow:
- Lossy → lossy only when necessary – for example for a device that cannot play the source format. In that case, the target bitrate should be set high to minimise the additional loss.
- Archive losslessly. If you want to preserve your CD collection or recordings long-term, convert them with WAV → FLAC into space-saving, lossless FLAC. From that FLAC master, any lossy file you need can be produced later – for example with FLAC → MP3 for the car stereo or FLAC → M4A for Apple devices.
The reverse also holds: MP3 → FLAC makes a file losslessly storable, but not better. The MP3 artefacts are merely preserved bit-exactly. At most, this is useful to avoid further re-encoding within a processing chain.
Recommendations by use case
- Speech (podcasts, audiobooks, dictation): 64–96 kbps MP3 mono or 48 kbps Opus is plenty. At 64 kbps, a 10-hour audiobook takes ~288 MB – instead of 1.4 GB at 320 kbps.
- Music on the go / Bluetooth: 128–160 kbps AAC/Opus or 160–192 kbps MP3 (VBR V2). Anything above that is lost in Bluetooth re-compression anyway.
- Music library for serious listening: 256 kbps AAC or MP3 VBR V0 (~245 kbps). This is practically transparent at half the storage cost of FLAC.
- Archive / master / editing: Lossless (FLAC or WAV). Every later edit or re-encode then starts without accumulated damage.
- 320 kbps CBR MP3: Honestly, rarely the rational choice. VBR V0 sounds just as good at ~25% smaller size, and AAC/Opus are more efficient. 320 kbps mainly makes sense where MP3 compatibility is mandatory and storage does not matter – which is why DJ pools often deliver 320s.
FAQ
Is 320 kbps MP3 the same as CD quality?
No. Even 320 kbps is lossy – the file contains only about 23% of a CD's data. In blind tests, the difference is imperceptible to almost all listeners. Technically, however, it is not identical. Bit-exact CD quality only comes from lossless formats such as FLAC, ALAC or WAV.
Can I restore quality from a 128 kbps file?
No. The information discarded during encoding is gone permanently. Upconverting to 320 kbps or FLAC only makes the file bigger. For better quality, a new source is required (CD, purchased download, lossless stream).
Which is better: 256 kbps AAC or 320 kbps MP3?
At least equal in listening tests, and AAC often comes out ahead – the codec is simply more modern and efficient. The bigger number on the MP3 side is misleading. More important than the bitrate, however, is a good encoder for either format (Apple AAC or LAME, respectively).
What bitrate does WhatsApp use for voice messages?
Voice messages are encoded by WhatsApp as Opus at very low bitrates (in the region of 16–32 kbps). For speech, that is enough – a good example of how efficient modern codecs are on suitable material.
Why do two 128 kbps files sound differently good?
Because the codec, the encoder version and the file's history all matter. A file freshly encoded with a current LAME build sounds noticeably better than one produced by an encoder from the 2000s – and better still than a file that has already been transcoded several times.
Conclusion
Bitrate is a budget, not a seal of quality. 128 kbps MP3 involves audible compromises. From roughly 190–245 kbps (VBR), MP3 becomes transparent for most listeners. Modern codecs like AAC and Opus achieve the same at 160 and 128 kbps respectively. The step up to 320 kbps costs a lot of storage for a difference that hardly anyone can demonstrate in a blind test. The rational strategy: archive losslessly, export to lossy as needed, and avoid repeated transcoding.
Convert audio with wandlio
For practical use, wandlio offers, among others, these audio converters:
- FLAC → MP3: produce space-saving files for on the go from your lossless archive
- WAV → FLAC: shrink recordings losslessly to roughly half their size
- MP3 → M4A and M4A → MP3: move between the two most widespread lossy ecosystems
- FLAC → M4A: prepare lossless masters for Apple devices
No registration is required for conversion, and uploaded files are deleted after processing.