Guides › Photos & files
128, 192 or 320 kbps? What MP3 bitrate changes, measured
When you pull the audio out of a video, the converter asks for a bitrate: 128, 192, 256 or 320 kbps. Higher sounds better, but the files get bigger. We took the guesswork out by running the same one-minute recording through our video to MP3 converter at every setting and measuring the result.
Short answer: size is exactly proportional to bitrate: about 0.96 MB per minute at 128 kbps and 2.4 MB at 320 kbps. The clearest difference we measured was bandwidth: 128 kbps cut everything above 16.7 kHz, while 192 kbps kept sound up to 18.7 kHz and 320 kbps up to 20.1 kHz. For speech, 128 kbps is plenty; for music, 192 kbps is a safe default.
How we tested
We generated one minute of stereo white noise at 44.1 kHz, the sample rate of CDs, and saved it as an uncompressed WAV file of 10.6 MB. White noise is the hardest possible signal for an audio encoder, because it contains equal energy at every frequency and nothing predictable to exploit, so it shows the encoder's limits clearly. We converted it in Chromium with our tool, which uses lamejs, a JavaScript port of the LAME MP3 encoder, and then decoded each MP3 and measured its frequency content.
File size
| Bitrate | Size for 1 minute | Per hour | Compared with WAV |
|---|---|---|---|
| WAV (uncompressed) | 10.6 MB | 635 MB | 100% |
| 128 kbps | 0.96 MB | 58 MB | 9% |
| 192 kbps | 1.44 MB | 86 MB | 14% |
| 256 kbps | 1.92 MB | 115 MB | 18% |
| 320 kbps | 2.40 MB | 144 MB | 23% |
Our converter uses a constant bitrate, so the size depends only on duration: bitrate × seconds ÷ 8 gives bytes. The content of the audio makes no difference. (The tool displays sizes in binary megabytes, so a 0.96 MB file appears as 0.9 MB.) Each conversion of the one-minute file took between 4 and 5 seconds on our test machine.
What is lost: the high frequencies first
To fit audio into fewer bits, the encoder first removes the highest frequencies, where hearing is least sensitive, and then simplifies detail that is masked by louder sounds. The first effect is easy to measure. We found the highest frequency that remained within 20 dB of the level in the 2 to 8 kHz range:
| Version | Highest frequency kept | Level at 18 kHz |
|---|---|---|
| Original WAV | 22.0 kHz (the format's limit) | unchanged |
| 128 kbps | 16.7 kHz | removed (−83 dB) |
| 192 kbps | 18.7 kHz | unchanged |
| 256 kbps | 19.4 kHz | unchanged |
| 320 kbps | 20.1 kHz | unchanged |
Whether you can hear the difference depends on your ears as much as on the file. Sensitivity to high frequencies declines steadily with age, as the international standard on age-related hearing thresholds documents, and many adults hear little above 16 kHz. Cymbals, breath sounds and room ambience live partly up there, so trained listeners with good headphones may notice a duller top end at 128 kbps on music.
Bandwidth is not the only difference. At low bitrates, encoders can also produce smearing before sharp sounds such as castanets or a snare drum ("pre-echo") and a swirling quality on applause or rain. Those are harder to measure with a single number, which is why listening tests remain the final judge.
The same test on an orchestral recording
White noise is a worst case, so we repeated the test with music: the last minute (1:34 to 2:34) of Grieg's "In the Hall of the Mountain King" from a recording that the Musopen project released into the public domain (linked in the references at the end of this page), a passage with the full orchestra and cymbals. We saved it as a 44.1 kHz, 16-bit stereo WAV of 10.6 MB, chose it on the converter's page in Chrome 153 on a four-core 2.1 GHz Linux machine, converted it at each setting, downloaded each MP3 and compared it with the original.
| Bitrate | MP3 size | Time to convert | Sound kept up to | 16–20 kHz level against the original |
|---|---|---|---|---|
| 128 kbps | 960,470 bytes | 6.0 s | 16.75 kHz | −16.7 dB |
| 192 kbps | 1,440,705 bytes | 5.8 s | 18.75 kHz | −1.8 dB |
| 256 kbps | 1,920,940 bytes | 5.8 s | 19.5 kHz | −0.4 dB |
| 320 kbps | 2,401,175 bytes | 5.1 s | 20.25 kHz | 0.0 dB |
- The sizes matched the noise test. Music or noise, one minute at 128 kbps came to 0.96 MB and at 320 kbps to 2.40 MB. With a constant bitrate the content does not change the size.
- The cut-offs were in the same places. Each setting stopped passing sound at the same point as with noise, within the 250 Hz steps used here: just under 17 kHz at 128 kbps and just over 20 kHz at 320 kbps.
- In real music, little sits up there. In this passage only about 0.3% of the sound energy was above 16 kHz, cymbals included. What 128 kbps removes is real, but it is a small part of the recording, which is one reason the difference can be hard to hear. Between 12 and 16 kHz, every setting kept the level within 0.1 dB of the original.
- Each step up is closer to the source. After lining each MP3 up with the original and subtracting, the leftover difference was 14 dB below the music at 128 kbps, 22 dB at 192, 27 dB at 256 and 31 dB at 320. This figure does not measure what you hear, because the encoder deliberately puts its errors where louder sounds mask them, but it shows the direction.
You can download the numbers and check them. Both files are .csv and open in any spreadsheet; each starts with a header row.
- Results for the orchestral recording: one row per bitrate with the MP3 size in bytes, the time from choosing the file to "Done", the size the tool displayed, the frequency up to which sound was kept (in 250 Hz steps), the share of the original's energy above 16 kHz, the level of the 12–16 kHz and 16–20 kHz ranges against the original, and the difference figure.
- Level by frequency band: for a 250 Hz band at each whole kilohertz from 10 to 22 kHz (the last one ends at 22.05 kHz, the limit of the format), how far each MP3 was above or below the original in decibels.
Which bitrate to choose
| Audio | Suggested bitrate | Reason |
|---|---|---|
| Lectures, podcasts, voice memos | 128 kbps | Speech contains little above 8 kHz; smaller files |
| Music for phones and casual listening | 192 kbps (our default) | Keeps nearly all audible bandwidth at 60% of the 320 kbps size |
| Music you will edit or keep long-term | 256 to 320 kbps | More headroom if the file is processed again later |
| Source is already a low-quality stream | Match the source, often 128 kbps | A higher setting cannot restore what is missing |
Converting from video does not improve quality
Most videos already store their sound in a compressed format, usually AAC at 128 to 256 kbps. Converting it to a 320 kbps MP3 does not bring back anything the original encoder removed; it only makes a bigger file containing the same audio plus a second round of encoding. A good rule is to pick a bitrate equal to or slightly higher than the source, not the maximum by default. If you need the audio for professional editing, keep the original video file as your master.
Privacy and limits of a browser converter
Our converter decodes and encodes everything in your browser; the file is not uploaded. That makes it suitable for private recordings, but it also means the work happens on your device: very long videos need enough memory to hold the decoded audio, and the browser must be able to decode the video's audio format. If a file fails, try another browser or a shorter clip.
Common questions
Is VBR (variable bitrate) better?
Variable bitrate spends more bits on complex passages and fewer on simple ones, so it often gives similar quality in a smaller file. Our converter uses a constant bitrate because the size is predictable and every player handles it.
Why does the MP3 sound quieter or louder than the video?
The converter does not change the volume. A difference usually comes from the video player applying its own volume normalisation.
Is MP3 still worth using?
Newer formats such as AAC and Opus give better quality at the same bitrate, but MP3 plays on virtually every device and car stereo ever made, which is why it remains the safest choice for sharing.
References
- LAME MP3 encoder documentation: presets and lowpass filters
- ISO 7029:2017, Statistical distribution of hearing thresholds related to age and gender
- ISO/IEC 11172-3: MPEG-1 Audio Layer III
- MDN: Web audio codec guide
- Internet Archive: Musopen Collection as FLAC (public-domain recordings)
Written and checked by Ria Kim for GlowHalo; figures without a source link come from the steps shown on this page. Companion tool: Video to MP3 Converter. Spotted an error? Write to help@glowhalo.org. Chrome and Chromium are trademarks of their owners; GlowHalo is not affiliated with them.