Compress MP3 Files: Lower the Bitrate in Your Browser
To compress MP3 files you change one thing: an MP3's size is set almost entirely by bitrate and duration, the duration is fixed, so the bitrate is what you lower. Drop MP3, WAV or M4A files into the tool below and pick a target bitrate, or just type “smaller than this many MB”. The tool works out each file's estimated size from the formula before it encodes anything. Voice can go to mono, which buys roughly the same clarity for half the bitrate, and at very low bitrates the tool also lowers the sample rate so the few bits left go to the speech band. It all runs in your browser; nothing is uploaded.
- Target bitrate or size
- Voice to mono
- Estimated size first
- Nothing uploaded
Updated Nov 6, 2026Fact-checked Nov 6, 2026
Compress MP3
ON DEVICECompress MP3: drop files, pick a bitrate or a target size
The list shows every file's current size and its estimated size after compression; files that wouldn't get smaller are flagged so you keep the original. Several files come as one ZIP.
How to compress MP3 files in four steps
First decide whether you want “sounds good enough” or “smaller than a limit”. For the first, choose a bitrate; for the second, type the size and let the tool work the bitrate out.
- 01
Add the files
Drop one or more MP3, WAV or M4A files, up to 50 per run. The list shows each file's duration and current size. Each file can be up to 200 MB and 60 minutes, half that on phones.
- 02
Pick a target bitrate, or type a target size
Bitrate mode offers 24 to 192 kbps (64 by default). Size mode takes a limit (5 MB by default, with 1 MB counted as 1,048,576 bytes) and picks the highest standard bitrate that fits under it, leaving about 2% for tags. If even the lowest step won't fit, the tool says so up front.
- 03
Switch voice to mono
Mono is ticked by default; keep it for dubbing, voiceover and audiobooks: for the same size, mono gets twice the bitrate per channel. Keep stereo only for files with stereo music or a real left-right sound field.
Tip · At very low bitrates the tool drops the sample rate to 22.05 or 16 kHz on purpose, because it sounds better, not because something went wrong.
- 04
Check the estimate, then compress
The estimate uses the constant-bitrate formula; the real file only adds a few KB of tags. When the estimate is about the same as the original, the tool says compressing won't make it smaller and hands you the original file instead of encoding it again. Confirm and compress; results carry the bitrate in their name (such as -64k.mp3), and several files can come as one ZIP.
How big will a compressed MP3 be? Bitrate × duration
A constant-bitrate (CBR) MP3 stores a fixed number of bits every second, so size = bitrate × duration ÷ 8. You can't change the duration; how small the file gets depends on how low the bitrate can go.
| Bitrate | 2-minute episode | 15-minute compilation | 60-minute audiobook chapter |
|---|---|---|---|
| 128 kbps | 1.92 MB | 14.4 MB | 57.6 MB |
| 96 kbps | 1.44 MB | 10.8 MB | 43.2 MB |
| 64 kbps | 0.96 MB | 7.2 MB | 28.8 MB |
| 32 kbps | 0.48 MB | 3.6 MB | 14.4 MB |
Working back from a target size
Turn the formula round: highest bitrate = target size (MB) × 8,000 ÷ seconds. Say a 15-minute (900-second) compilation must stay under 10 MB. The tool counts 1 MB as 1,048,576 bytes and keeps 2% for tags: 10 × 1,048,576 × 0.98 × 8 ÷ 900 ÷ 1,000 ≈ 91 kbps. MP3 bitrates come in fixed steps, and the highest one not above 91 is 80 kbps, which gives 9,000,000 bytes, shown in the tool as about 8.58 MB. Ask for 1 MB instead and the answer is under 10 kbps, below the lowest step; the tool tells you it can't be done rather than squeezing out a file nobody can follow.
Why the sample rate drops at low bitrates
Sample rate sets the highest frequency a file can hold (about half the sample rate); bitrate sets how much data per second describes those frequencies. Keep 44.1 kHz at a very low bitrate and a small budget is spread across the whole 0–22 kHz band, described coarsely everywhere, which you hear as watery, metallic artefacts. Drop to 22.05 kHz and only the range below about 11 kHz is kept, so the data goes where speech actually lives and the same bitrate sounds cleaner. Apple Podcasts' audio requirements describe the same idea: when bitrate is constrained, reduce the sample rate from 48 to 24 kHz, reduce stereo to mono, or both. This tool's rules are below.
| Target bitrate | Mono output sample rate | Stereo output sample rate |
|---|---|---|
| 96 kbps and above | Kept at 44.1 / 48 kHz | Kept at 44.1 / 48 kHz |
| 64–80 kbps | Kept at 44.1 / 48 kHz | 22.05 kHz |
| 48–56 kbps | 22.05 kHz | 22.05 kHz |
| 32–40 kbps | 22.05 kHz | 16 kHz |
| Below 32 kbps | 16 kHz | 16 kHz |
Does compressing an MP3 lose quality? How low voice can go
It does lose quality; the question is how much is acceptable. Voice compresses far better than music: a narrow band and little dense high-frequency detail mean dialogue stays clear at bitrates music would never survive.
Published numbers to lean on
| Channels | Recommended at 44.1 / 48 kHz | Recommended at 22.05 / 24 kHz |
|---|---|---|
| Mono | 64–128 kbps | 40–80 kbps |
| Stereo | 128–256 kbps | 80–160 kbps |
The table says two things. Voice in mono can take half the bitrate and stay inside the recommendation, and at 22.05 / 24 kHz even 40–80 kbps mono counts as fine. AI Drama Voice exports dubs at 44.1 kHz mono 128 kbps, the top of the mono range; at 64 kbps mono the file halves and is still inside it. Around 32 kbps speech is still understandable, but sibilants (s, sh, ch) turn mushy and any music underneath suffers badly.
Compressing a file that is already MP3: a second loss
Decoding an MP3 and encoding it again strips detail from sound that has already lost some; this is generation loss. With the original WAV at hand, compressing straight from it to your target is cleaner than going to 128 kbps first and then to 64. And “compressing” a 64 kbps file at 96 kbps only makes it bigger at the same quality, which is why the tool flags files that wouldn't get smaller.
Will my subtitles still line up?
Yes. Compression changes the data per second, not the duration, so every line stays where it was; re-encoding shifts things by a few tens of milliseconds at most, which no one notices on subtitles. The episode's SRT still works and needs no SRT time shift.
Test one file first
Before a batch, compress the hardest file (the one with music or the most heated scene), listen on headphones, and only then run the rest at that bitrate.
When dubbed audio is worth compressing
AI Drama Voice exports each episode as mono 128 kbps, under 1 MB per minute, so most of the time there is nothing to compress. These are the cases where it pays.
Compilations get big
Join a season into one file with merge MP3 files and half an hour is over 20 MB. For a review copy, 64 kbps mono is plenty.
Upload forms have limits
Platforms, chat apps and email services each set their own attachment limit, so go by what the other side states. Type that limit into target size and the tool picks the highest bitrate that fits.
Recordings and WAV deliverables
WAV from a recorder or an editor takes 5–10 MB a minute; compressing straight to MP3 is the quickest fix. For more format options, use the WAV to MP3 converter.
Match loudness first
Compression doesn't change volume. To even out episodes, use normalize audio volume first and compress last, so you encode one time fewer.
Keep the original
Compression can't be undone. Keep the original MP3 or WAV and export from it when you need a higher bitrate later; never convert a compressed file back up.
Tag after compressing
Compression re-encodes, so the new file doesn't carry the source file's ID3 tag, including the TXXX:AIGC label our exports write. To give compressed episodes titles, an album and cover art, use the MP3 tag editor, which edits tags without encoding again; add the visible AI label your platform asks for when you publish.
Frequently asked questions
Yes; to compress MP3 audio is to describe it with less data. Voice is forgiving: 64 kbps mono is still inside Apple Podcasts' recommended range and dialogue holds up well. The lower the bitrate and the more music in the mix, the more you hear it.
About 64 kbps mono is a safe floor for voice, and 40–48 kbps works once the sample rate drops to 22.05 kHz; below 32 kbps it is understandable but muddy. In size terms, 64 kbps is about 0.48 MB per minute, so a two-minute dubbed episode is under 1 MB.
Yes. Choose target size and type the limit; the tool works back from bitrate × duration ÷ 8, picks the highest standard bitrate that fits and shows the estimate first. When the file is too long for the target even at the lowest step, it tells you.
At the same bitrate, not at all: both are the same size. The gain is that mono spends the whole bitrate on one channel, so voice in mono can take half the bitrate for similar clarity. Stereo only matters for music and ambience with a real left-right field.
Yes. Drop in several files and they share one bitrate or one target size, then download together as a ZIP. Each file's estimated size is shown before anything is encoded.
Yes. Compression lowers the bitrate without changing the duration, so every line keeps its position and the existing SRT still works; re-encoding moves things by a few tens of milliseconds at most.
No. Audio is decoded and re-encoded in your browser, so everything you do to compress MP3 files stays on your device. No AI, no credits and no sign-in.
Files compressed. Let us dub the next one. Small files, same clear voices.
Paste a Chinese script or novel promo: AI casts the characters, dubs it line by line and keeps the same cast across 8 languages, with MP3 and SRT exports. Cast analysis is free.
Billed by characters generated · Failed runs refunded automatically · Alipay, WeChat Pay and cards