WAV MP3 Converter

Shrink WAV to MP3, or unpack MP3 to the WAV spec you need, without uploading anything

The WAV MP3 Converter turns WAV files into MP3 and MP3 files into WAV. Drop a file and it lands in the matching list (WAV → MP3 or MP3 → WAV), then a frequency spectrum shows which pitches the file actually contains and how loud they are. When you pick a quality or a sample rate, the band that will be lost is hatched on the spectrum, and the exact size of the result is shown in bytes before you convert.

The example WAV on the first screen (12 seconds of synthetic guitar, 44.1 kHz, 16-bit stereo) is 2,116,844 bytes. Converted to a 192 kbps MP3 it becomes 289,018 bytes, which is 86% smaller, and the estimate on screen matches the saved file exactly. All of the work happens inside this browser tab, and your files are never sent to a server.

When to convert WAV to MP3, and MP3 to WAV

WAV stores sound uncompressed, which makes it easy to edit but large. MP3 drops the parts of the sound that people notice least to get a much smaller file, and what it drops cannot be recovered. The two directions therefore serve different jobs.

• Sending a recording (WAV → MP3): a 44.1 kHz stereo WAV from a recorder or a phone's voice memo app runs to about 10.6 MB per minute, which quickly hits email and chat attachment limits. The same minute as a 128 kbps MP3 is 960,470 bytes (about 0.96 MB).
• Feeding something that only takes MP3 (WAV → MP3): some car stereos, older MP3 players and upload forms accept nothing else. Renaming the extension does not help, because the file inside is still WAV and is likely to be rejected.
• Opening audio in an editor (MP3 → WAV): uncompressed WAV is easier to cut and splice, and some editors and hardware read only WAV. When you save the finished edit as MP3 again, the sound is compressed a second time.
• Meeting a required spec (MP3 → WAV): when the receiving side fixes the format, as a phone system, a speech recognition service or a video editor may (8 kHz, 16 kHz or 48 kHz, 16-bit, mono), pick the sample rate, bit depth and channels here and the WAV comes out in that format.

How to use it

1. Add files: press 'Add files' or drop files anywhere on the workspace. WAV files go to the WAV → MP3 list and MP3 files to the MP3 → WAV list, up to 20 per list at once.
2. Look at the spectrum: the average level of the selected file is drawn as a grey area. With several files, click a row to see another file's spectrum.
3. Choose settings: for WAV → MP3, pick a quality (64 to 320 kbps) and channels. For MP3 → WAV, pick the sample rate (original, 48, 44.1, 22.05, 16 or 8 kHz), bit depth (16 or 24-bit) and channels. Each button shows the size you would get, and the hatching moves to match your choice.
4. Convert: press 'Convert to MP3' or 'Convert to WAV' and the files in the list are converted one after another. The result bar fills in step with real progress, and 'Stop' ends the run early.
5. Listen and save: switch between 'Original' and the result to compare the same moment. While you listen, an orange line follows the spectrum of what is playing. A single file saves directly; with two or more you can download one ZIP or save them one by one from the list.

Reading the spectrum: the hatched band is what the result loses

The horizontal axis is pitch (frequency), from 0 Hz up to half the file's sample rate (24 kHz at most). Digital audio can only hold sounds up to half its sample rate, so a 44.1 kHz file is drawn up to 22.05 kHz. The vertical axis is level in dB. The curve averages 48 evenly spread sections of the file (32 for MP3), each cut into 4,096 points. The calculation is the same one the browser's Web Audio AnalyserNode uses, so the orange line you see during playback and the grey area share one scale.

• Hatching in WAV → MP3: LAME, the MP3 encoder used here, throws away everything above a cutoff frequency that depends on the bitrate and spends the saved bits on the lower range. The hatching starts at that cutoff.
• Hatching in MP3 → WAV: a WAV cannot hold anything above half its sample rate. Choose 8 kHz and everything above 4 kHz is hatched. A higher sample rate than the original shows no hatching, but it does not create sound the MP3 never had.
• The 'empty above here' dotted line: the frequency above which only sound more than 70 dB below the loudest band remains. For an MP3 it usually sits between 16 and 20 kHz, which is the band that was already discarded when that MP3 was made. If the dotted line is to the left of the hatching, you lose almost nothing real with that setting.

A mono file has one channel, so the same bitrate can cover a wider range, and LAME raises the cutoff by a factor of 1.5 (capped at 20.5 kHz). Converting white noise with this encoder, 64 kbps mono kept sound up to about 16.7 kHz and 128 kbps stereo up to about 16.6 to 17.0 kHz (measured 2026-10-20). People hear up to about 20 kHz, so a setting that drops everything above 11 kHz, such as 64 kbps stereo, dulls the very top of the sound, for example the sparkle of cymbals.

How the size is worked out

An uncompressed PCM WAV is seconds × sample rate × channels × bit depth ÷ 8 bytes plus a 44-byte header. A constant-bitrate MP3 is built from frames of 1,152 samples, and with this encoder the size is ⌊(⌈samples ÷ 1,152⌉ + 1) × 144,000 × kbps ÷ sample rate⌋ bytes. The extra 1 is the last frame the encoder flushes at the end. So an MP3's size depends only on its length and bitrate, never on the sound itself, which is why the estimate on screen and the saved file match to the byte.

MP3 → WAV goes the other way and the file grows. The table below shows the WAV size for a 3-minute 44.1 kHz stereo MP3 at several settings. The sound does not get better in proportion to the extra bytes; it is the same content from the MP3 in a bigger container.

Unpacking MP3 to WAV: length, sample rate and bit depth

An MP3 encoder adds a short delay before the sound and pads the last frame with empty samples. Files made with LAME record both counts in the info tag of the first frame. This tool reads them and trims both ends, so the WAV has the same length as the original recording, with the same sample count as ffmpeg gives. An MP3 without that tag, including the MP3s this tool makes, has no known delay, so every frame is decoded as is. The example MP3 therefore comes out 1,872 samples (about 0.04 seconds) longer than the original.

• Sample rate: 'Original' keeps the MP3's own rate. Any other value resamples. The filter starts cutting at about 94% of half the new rate, so sound more than 3% above that half is reduced by over 90 dB instead of folding back into lower pitches. In exchange, an 8 kHz WAV keeps an even response up to about 3.4 kHz, and sound close to 4 kHz comes out slightly weaker (measured with this tool's resampler on 2026-10-20).
• Bit depth: 16-bit is the default almost every program reads. 24-bit is 1.5 times the size; it only avoids the tiny rounding error of reducing the decoded sound to 16-bit and does not bring back anything the MP3 lost.
• Format: the result is a standard PCM WAV with a 44-byte header. Because RIFF stores sizes in 32 bits, a WAV over 4 GiB is not created. At 48 kHz, 16-bit stereo that limit is about 6 hours 12 minutes.

Shrinking WAV to MP3: sample rate, channels and WAV types

• Sample rate: WAV files at 32, 44.1 and 48 kHz, the rates MP3 (MPEG-1) uses, are kept as they are. A 96 kHz studio recording goes to 48 kHz, 88.2 kHz to 44.1 kHz, 8 and 16 kHz recordings to 32 kHz, and 22.05 kHz to 44.1 kHz. LAME on its own drops 64 kbps stereo to 24 kHz, but this tool fixes the rates above so that the file plays everywhere.
• Channels: a constant-bitrate MP3 is the same size in mono, but it can cover a wider band. If the left and right sides are identical, as with a single voice microphone, mono is the better choice. A WAV with three or more channels is mixed down to stereo; the centre and rear channels are added to left and right at 0.707 times the level of the front channels, the low-frequency effects channel (LFE) is left out, and the whole mix is lowered so it does not clip.
• WAV types: 8, 16, 24 and 32-bit integer PCM, 32 and 64-bit float PCM, the extensible format with channel information (WAVE_FORMAT_EXTENSIBLE) and RF64 beyond 4 GiB are all read. The file is read in pieces and encoded as it goes, so there is no size limit. Compressed WAV such as ADPCM or μ-law is converted only when the browser can decode it, and because that must be decoded all at once, only up to 20 minutes.

Common mistakes

• Only renaming the extension: changing .wav to .mp3 leaves the size unchanged and the contents still WAV. Depending on where you send it, it may not play or may be rejected.
• Converting MP3 to WAV to recover quality: the file only gets bigger, and the empty band stays empty. You can see this because the dotted line on the spectrum does not move.
• Assuming 320 kbps equals the original: 320 kbps still cuts everything above 20.5 kHz and compresses the rest. Keep the original as a WAV if you plan to edit it.
• Expecting mono to halve an MP3: at a constant bitrate the size is the same. To make the file smaller, lower the bitrate.
• Dropping to 8 kHz and then going back up to 44.1 kHz: sound above 4 kHz is gone for good. When you need a different spec, convert again from the first MP3.

Limits to know about

• Tags such as title and artist, and cover art, are not carried over to the result.
• There is no trimming, loudness matching or noise removal. To keep only part of a recording, use an audio cutter first.
• MP3 is created only at a constant bitrate (CBR), and WAV only as 16 or 24-bit PCM. μ-law, A-law and 32-bit float WAV are not created.
• In a browser without a built-in MP3 decoder (WebCodecs), the MP3 must be decoded all at once, so only up to 20 minutes can be turned into WAV. The browser's own resampler is used then, so the sample count can differ from the expected value by one.
• Refreshing or closing the page discards the results. Save the files you need first.

As of The standards and browser documentation were checked on 2026-10-20. The size tables, example results and low-pass measurements come from this tool's encoder, run in Chrome on the same day.

Sources

• 01MDN — Digital audio conceptsHalf the sample rate (the Nyquist frequency) is the highest sound that can be held; CD audio is 44.1 kHz, 16-bit stereo; people hear up to 20,000 Hz
• 02MDN — Web audio codec guideMP3 bitrates of 32 to 320 kbps; sample rates of 32, 44.1 and 48 kHz for MPEG-1 and 16, 22.05 and 24 kHz for MPEG-2
• 03LAME — libmp3lame/lame.cLow-pass frequency by bitrate (lowpass_by_bitrate), with mono scaled by 1.5
• 04LAME — Technical FAQOne MP3 frame is 1,152 samples, padding in the last frame, decoder delay. LAME is licensed under the LGPL
• 05LAME — Mp3 info tag specificationsThe encoder delay and padding recorded in the info tag
• 06Microsoft — WAVEFORMATEXTENSIBLEFormat information for WAV with more than two channels or higher resolution than WAVEFORMATEX, and the channel mask that maps channels to speaker positions
• 07EBU Tech 3306 — RF64RIFF/WAVE addresses only up to 4,294,967,296 bytes (4 GB) with 32 bits, so RF64 defines 64-bit addressing
• 08W3C — Web Audio API: AnalyserNodeFrequency data is computed with a Blackman window and converted to dB with 20 log10
• 09MDN — AudioDecoderThe browser's built-in decoder (WebCodecs) and which browsers support it
• 10wasm-media-encoders — LAME MP3 encoder for WebAssemblyMIT license (the LAME encoder itself is LGPL). This tool uses version 0.7.0 as is

Frequently asked questions

Tools to use with it