Audio File Formats Explained: MP3, WAV, AAC, FLAC, OGG

Audio formats generate more confident misinformation than almost any other technical subject. People insist that MP3 destroys music, that converting a file to WAV will improve it, that a higher sample rate always sounds better, and that a 320 kilobit file is worth four times the space of a 128 kilobit one under every circumstance. Some of that is true, some of it is precisely backwards, and most of it depends entirely on what the file is for. This guide explains what actually happens when audio is compressed, what each format was designed to do, what bitrate and sample rate really control, and how to pick the right format for recording, sharing, publishing, and archiving without guessing.

How Audio Becomes a File

Sound is a continuous wave of pressure moving through air. A computer cannot store a continuous anything, so the first thing a recording device does is measure that wave at fixed intervals and write down the numbers. This is sampling, and it is the foundation of every digital audio format in existence.

Two numbers define how that measurement is done. The sample rate is how many measurements are taken each second, commonly 44,100 for music or 48,000 for video work. The bit depth is how precisely each measurement is recorded, commonly 16 bits, giving 65,536 possible levels, or 24 bits, giving over sixteen million. Together they produce a stream of raw numbers, and the size of that stream is entirely predictable: sample rate multiplied by bit depth multiplied by the number of channels.

Run the arithmetic on CD quality stereo and the result is instructive. That is 44,100 samples per second, times 16 bits, times 2 channels, which comes to 1,411,200 bits per second, or about 10 megabytes per minute. An hour of audio is roughly 600 megabytes. A modest music collection stored that way would fill a hard drive quickly, and streaming it over a phone connection would be painful.

That number is why audio compression exists, and why it split into two completely different philosophies. One reduces size by discarding information deliberately. The other reduces size by finding patterns without discarding anything. Understanding which one you are using, and when each is appropriate, resolves nearly every audio format question there is.

Lossy vs Lossless: Where the Data Goes

The division between lossy and lossless is the single most important concept in audio formats, and it is far more intuitive than the terminology suggests.

How lossy compression works

Lossy encoders exploit the limitations of human hearing, a field called psychoacoustics. Several of these limitations are surprisingly large. You cannot hear frequencies above roughly 20,000 hertz, and most adults lose the top of that range with age, so anything up there can be deleted with no consequence. A loud sound masks a quieter sound at a similar frequency occurring at the same moment, so the quieter one can be discarded because you were never going to hear it anyway. A sudden loud sound also masks quieter sounds immediately before and after it, a phenomenon called temporal masking.

A lossy encoder models all of this, decides which parts of the signal are perceptually irrelevant, and throws them away permanently. What remains is stored efficiently. The result is a file roughly one tenth the size of the original that most people, on most equipment, cannot distinguish from it.

The word permanently is the important one. The discarded data is not compressed or hidden. It is gone. No conversion, no upscaling, and no processing will bring it back, which is the source of the most persistent misunderstanding in this entire subject.

How lossless compression works

Lossless audio compression works like a ZIP file. It looks for redundancy and patterns in the numbers and encodes them more efficiently, but every original value can be reconstructed exactly. Decompress a FLAC file and you get back a bit for bit identical copy of the WAV it came from.

Because it cannot discard anything, the savings are far more modest: typically forty to sixty percent, depending on the material. Dense, loud music compresses least, quiet acoustic recordings compress most. A ten megabyte WAV might become a five megabyte FLAC, whereas the same audio as a 192 kilobit MP3 would be closer to one megabyte.

Why transparency matters more than numbers

Audio people use the word transparent to mean a compressed file that cannot be reliably distinguished from the original in a properly blind test. Transparency is the practical goal, not maximum bitrate, because past the point of transparency you are spending storage on something nobody can hear.

Where transparency sits depends on the content and the listener. For most people, most music, and most equipment, a well encoded 192 kilobit MP3 is very close, and 256 or 320 is effectively there. Modern codecs such as AAC and Opus reach transparency at noticeably lower bitrates, because they are simply better at deciding what to keep.

The Formats, One by One

With the lossy and lossless split understood, each format becomes a straightforward set of trade offs rather than a mystery.

MP3, the universal survivor

MP3 was finalised in 1993 and became the format that made digital music portable. By modern standards its compression is dated, and at the same bitrate both AAC and Opus sound better. It has genuine technical limitations, including a small unavoidable gap at the start and end of every file, which is why gapless album playback was awkward for years.

None of that has dented its usefulness, because MP3 does one thing better than anything else: it plays absolutely everywhere. Car stereos from 2004, cheap MP3 players, every phone, every browser, every editing program, every embedded system. Its patents expired in 2017, so it is now completely free to use. When a file simply has to work with no questions asked, MP3 at 192 to 320 kilobits per second remains the right choice, and its slight inefficiency is a small price for certainty.

WAV, the raw working format

WAV is not really a compression format at all. It is a container holding uncompressed PCM audio, which means the raw sample values written straight to disk. Nothing is discarded and nothing is encoded, so there is no quality loss and no decoding overhead, but files are enormous at roughly ten megabytes per minute of stereo CD quality audio.

That makes WAV excellent for exactly one phase of the process: recording and editing. Every cut, fade, level change, and effect applied to a WAV file operates on the full original data, so nothing degrades as you work. It is a poor choice for distribution, and a mediocre choice for archiving, because it has weak and inconsistent support for metadata tags. You can store a WAV, but you will struggle to reliably store the artist, album, and track number alongside it.

AAC, the better MP3

AAC, Advanced Audio Coding, was designed as MP3's successor and succeeds at it. At any given bitrate it sounds better, with the advantage most pronounced at lower bitrates where MP3 struggles badly. A 128 kilobit AAC file is roughly comparable to a 192 kilobit MP3, which is a substantial saving.

AAC is the default audio format inside MP4 video, the standard for Apple Music and YouTube, and the audio track of essentially every video file you will encounter. Support is now near universal on anything made in the last fifteen years, though a handful of very old devices still only accept MP3. It normally appears with an .m4a extension when stored as pure audio.

FLAC, lossless done right

FLAC, the Free Lossless Audio Codec, is the correct answer for archiving. It compresses without discarding anything, producing files roughly half the size of WAV that decode to bit for bit identical audio. It supports full metadata tagging, embedded cover art, and per file checksums that let you verify a file has not become corrupted over the years, which genuinely matters for a collection you intend to keep for decades.

It is open and royalty free, and support is now broad across music players, phones, and media servers, though it is still absent from some car systems and older hardware. The standard approach is to archive in FLAC and generate lossy copies from it whenever a portable version is needed. ALAC, Apple Lossless, does the same job with better native Apple integration and slightly worse support elsewhere.

OGG and Opus, the open alternatives

Ogg Vorbis was created as a patent free alternative to MP3 and outperforms it comfortably, particularly at moderate bitrates. It became the standard audio format in video games, where the licensing freedom mattered a great deal, and it remains common in open source software.

Opus is its successor and is genuinely remarkable. It handles everything from very low bitrate speech to full quality music in a single codec, and at low bitrates it is far ahead of anything else. Opus at 64 kilobits per second sounds better than MP3 at 128. It is the audio codec behind most voice and video calling, and it pairs with VP9 and AV1 inside WebM.

The limitation for both is playback support outside browsers and software players. Standalone hardware often refuses them, which is why OGG files downloaded from games or open source projects so often need converting before they will play on a phone or in a car. An OGG to MP3 converter resolves that in one step.

The rest, briefly

M4A is a container rather than a codec, and usually holds AAC, though it can hold ALAC. WMA is Microsoft's format, effectively obsolete outside legacy systems. AIFF is Apple's uncompressed equivalent of WAV, with better metadata support. AC 3 and E AC 3 are the Dolby formats found on discs and in broadcast, which is why a downloaded film sometimes plays picture with no sound on a device that cannot decode Dolby. AMR is a very low bitrate speech codec from old mobile phones, worth converting immediately if you find one.

Bitrate, Sample Rate, and Bit Depth

Three numbers get quoted constantly and confused just as often. Each controls something genuinely different, and only one of them is usually worth adjusting.

Bitrate, the one that matters most

Bitrate is how much data is spent on each second of audio in a lossy file, and it is the primary quality control. More data means the encoder discards less, which means the result is closer to the original.

The useful reference points are these. At 64 kilobits per second, music sounds obviously degraded but speech is perfectly clear. At 128, most listeners hear artefacts on cymbals, applause, and reverb tails, but casual listening is fine. At 192, the difference from the source becomes hard to detect on typical equipment. At 256 and 320, it is effectively transparent for almost everyone. Above 320, MP3 simply cannot use the extra data productively.

Bitrate can also be constant or variable. Variable bitrate lets the encoder spend more data on complex passages and less on simple ones, producing better quality for the same average size. It is the better default in almost every case, and the only reason to use constant bitrate is compatibility with unusual old hardware or streaming systems that require a fixed rate.

Sample rate, which you should usually leave alone

Sample rate determines the highest frequency that can be recorded, and the relationship is fixed: the maximum reproducible frequency is exactly half the sample rate. At 44,100 hertz, the ceiling is 22,050 hertz, which is already above the limit of human hearing. That is not a coincidence, it is why CD audio uses that number.

This is the source of a persistent myth. Raising the sample rate to 96,000 or 192,000 does not make audio sound better to a listener, because the additional frequencies captured are inaudible. High sample rates have a legitimate purpose during production, where processing such as pitch shifting and time stretching benefits from the extra headroom, but for a finished file they add size and nothing else.

Use 44,100 for music and 48,000 for anything that will accompany video, since that is the video standard and matching it avoids an unnecessary resampling step. Never upsample a finished file. Converting a 44,100 hertz recording to 96,000 invents nothing and merely doubles the size.

Bit depth, which matters while recording

Bit depth controls the precision of each measurement, and in practice it controls the dynamic range, meaning the distance between the quietest sound the file can represent and the loudest. Sixteen bits gives roughly 96 decibels of range, which comfortably exceeds what any normal listening environment requires. Twenty four bits gives roughly 144 decibels.

The extra range is genuinely valuable while recording, because it means you can set levels conservatively, leave plenty of headroom to avoid clipping, and still have a clean signal after you raise the level in editing. It provides no benefit at all in a finished file that someone is going to listen to. Record at 24 bit, deliver at 16 bit.

Mono, Stereo, and Channel Choices

Channel count is a straightforward multiplier that is regularly ignored. Stereo is two channels and therefore roughly twice the data of mono. That cost is entirely justified for music, where the stereo image is part of the recording, and entirely wasted on a single speaker recorded on one microphone, where both channels contain the same signal.

A great deal of spoken word content is distributed in stereo purely because nobody changed the default, which doubles the file size for no benefit whatsoever. Converting an interview or a solo narration to mono halves the size with literally no loss of information, since there was no stereo information to begin with.

Joint stereo, which most encoders use by default, sits in between. It encodes the shared content of both channels once and only stores the differences, which is efficient and generally indistinguishable from full stereo. It is a sensible default and rarely needs changing.

Choosing a Format by Purpose

As with video, the destination answers the question. Work backwards from where the file is going and the choice becomes obvious.

Recording and editing

WAV at 24 bit, 48,000 hertz. Nothing is compressed, so nothing degrades across dozens of edits, and the headroom protects you from level mistakes. Convert only at the very end, once the audio is final.

Archiving music

FLAC. Identical to the source, roughly half the size of WAV, properly taggable, and verifiable with built in checksums. Every future format you might need can be generated from it without loss, which is the entire point of an archive.

Podcasts and spoken word

MP3 at 96 to 128 kilobits per second in mono. Voice sits in a narrow frequency range and compresses very efficiently, and mono halves the data with no loss for a single microphone. That works out at roughly 50 megabytes per hour, which downloads quickly on a poor connection and sounds entirely clean. MP3 rather than AAC here purely because podcast apps and directories have the longest compatibility tail of anything in audio.

Music for portable listening

MP3 at 256 or 320, or AAC at 192 to 256 if the devices support it. Both are transparent for practical purposes. Choose based on what has to play the file, not on which codec is technically superior.

Audio inside video

AAC at 128 to 192 kilobits per second. It is what MP4 expects, what every platform accepts, and what every device decodes in hardware. Using anything else in a video file invites the classic failure where the picture plays and the sound does not.

Web and streaming

Opus in an OGG or WebM container where you control the player, since it is dramatically more efficient at low bitrates. MP3 as a fallback where you do not.

Sending a file to someone

MP3, essentially always. Whatever they are opening it with will accept it. This is the one case where being technically outdated is an advantage.

Converting Between Formats Safely

Conversion is where good audio quietly becomes bad audio, almost always through one of a small number of avoidable mistakes.

The conversions that cost nothing

Lossless to lossless is free. WAV to FLAC, FLAC to WAV, FLAC to ALAC and back again all preserve the audio exactly, because nothing is being discarded at any stage. You can move between them as often as you like with no cumulative damage.

Lossless to lossy costs a single, controlled generation of loss, and that is expected and fine. This is the normal path: archive in FLAC, export MP3 or AAC copies whenever needed. Because the archive is intact, you can regenerate those copies at any quality, in any format, forever.

The conversions that cost something

Lossy to lossy is the one to avoid. Converting an MP3 to AAC, or a 320 kilobit MP3 to a 192 kilobit MP3, means the second encoder is analysing audio that has already had material removed. It makes its own decisions on a degraded signal, and the artefacts compound. Sometimes it is unavoidable, for instance when a device only accepts one format and all you have is another, but it should never be routine.

When you must do it, keep the bitrate the same or higher than the source. Going from 320 to 128 discards a great deal on top of what was already gone.

The conversion that does nothing

Lossy to lossless is the mistake that seems most sensible and helps least. Converting an MP3 to WAV or FLAC produces a file five to ten times larger containing exactly the same audio quality. The discarded information does not come back. All you have done is store the degraded version inefficiently.

There is one narrow, legitimate exception. If you need to perform serious editing on a file that only exists as an MP3, converting it to WAV first prevents further generation loss during the editing process itself. You are not improving the audio, you are stopping it getting worse while you work on it.

Trimming, merging, and levels

Plenty of audio work is not conversion at all. Cutting the silence and false starts off the front of a recording, joining several takes into one continuous file, or lifting a quiet recording to a usable level are all common tasks that are best handled before any final encoding, so that the lossy pass happens once at the end.

Trimming dead air with an audio trimmer is usually the first thing to do with any raw recording, since it shrinks the file and improves the listening experience at the same time. Joining segments with an audio merger handles multi part recordings. And when a recording came out too quiet, which happens constantly with phone recordings and conference calls, an audio volume booster fixes the level rather than leaving listeners to hunt for their volume control.

Getting audio out of video

A very common need is extracting the sound from a video, whether that is a recorded meeting, a lecture, an interview you filmed, or a track you want for transcription. There is no reason to keep a two hundred megabyte video file when only the audio matters, and pulling it out with a video to MP3 converter typically replaces it with something under five megabytes. Our companion guide to video file formats and codecs explains how the audio track sits inside the video container and why some files play picture with no sound.

Privacy while converting

Audio is frequently sensitive in ways video is not always. Recorded meetings, therapy sessions, legal proceedings, interviews given under conditions, and unreleased music all pass through converters routinely. The standard model for an online converter is to upload the file to a server, process it there, and let you download the result, which means a complete copy exists on somebody else's infrastructure under somebody else's policy.

Browser based tools avoid the issue completely by doing the work locally, using your own processor, with nothing transmitted. For anything confidential that is not a minor preference, it is the whole difference between a private operation and a disclosure.

Fixing Common Audio Problems

The file will not play on my device

Usually FLAC, OGG, or Opus meeting hardware that only understands MP3 and AAC. Convert to MP3 and the problem disappears. Keep the original, since the conversion is one way.

The video plays but there is no sound

The audio codec inside the video is not supported, most often AC 3 or E AC 3 from a disc rip, or Opus on an older device. The video needs its audio track re encoded to AAC.

The recording is far too quiet

Input levels were set too low at the time of recording. Raising the level afterwards works but also raises the noise floor, which is why the hiss becomes noticeable. It is recoverable, but the real fix is to record with levels peaking around minus six decibels next time.

The audio drifts out of sync with the video

Nearly always an old AVI file with variable bitrate audio in a container that cannot represent it correctly. Converting the whole file to MP4 with AAC audio rebuilds the timing properly.

The file is too large to upload

Check the three usual culprits in order: it may be in a lossless format when lossy would do, it may be stereo when mono would do, and the bitrate may simply be higher than the content needs. An hour of speech has no business being a 600 megabyte WAV. Running it through an audio compressor to a sensible bitrate typically cuts it by well over ninety percent with no audible difference on voice.

Music sounds thin or hollow after conversion

Classic symptom of a low bitrate lossy encode, or of a lossy to lossy conversion. Re encode from the original source at a higher bitrate. If the original is gone, the quality is gone with it, and no processing will restore it.

Frequently Asked Questions

What is the best audio format to use?

It depends on the stage. WAV while recording and editing, FLAC for archiving, MP3 at 192 to 320 for anything that must play everywhere, and AAC at 128 to 192 for audio inside video or for modern devices. There is no single winner, only the right tool for the phase.

Does converting MP3 to WAV improve quality?

No. Converting lossy to lossless cannot restore discarded information. You get a much larger file with exactly the same audio quality. The only reason to do it is to avoid further loss during editing.

Is 320 kbps really better than 128 kbps?

Yes, measurably, and often audibly on cymbals, applause, and reverb tails. Whether you personally can hear it depends on the material, the equipment, and the listening environment. At 192 the difference from the source is already hard to detect for most people.

What bitrate should a podcast use?

96 to 128 kilobits per second in mono. Voice compresses efficiently and mono halves the data with no loss for a single microphone, producing around 50 megabytes per hour that downloads quickly and sounds clean.

Should I archive in FLAC or WAV?

FLAC. The audio is identical, the files are roughly half the size, metadata tagging actually works, and built in checksums let you verify years later that nothing has become corrupted.

Does a higher sample rate sound better?

Not for listening. Sample rate sets the highest recordable frequency, and 44,100 hertz already covers the entire range of human hearing. Higher rates are useful during production but add only file size to a finished file.

Why is my audio file so large?

Almost always because it is uncompressed, it is stereo when it did not need to be, or the bitrate is far higher than the content requires. An hour of uncompressed stereo is around 600 megabytes. The same hour of speech as a mono 96 kilobit MP3 is about 43 megabytes and sounds no different.

Can I convert audio without uploading it anywhere?

Yes. Browsers can decode and encode audio locally, so a well built tool does the work on your own machine with nothing transmitted. For recorded meetings, interviews, and unreleased material that distinction matters considerably.

The Short Version

Audio compression comes in two flavours, and knowing which one you are holding answers nearly every question. Lossy formats such as MP3, AAC, and Opus permanently discard what you are unlikely to hear, making files roughly ten times smaller and perfect for listening and sharing. Lossless formats such as FLAC discard nothing, cost about half the size of raw WAV, and are the right way to keep anything you care about.

Record in WAV, archive in FLAC, share in MP3, and put AAC inside video. Bitrate is the number worth adjusting, sample rate and bit depth usually are not. Mono halves the size of spoken word for free. And the one rule that never bends: quality can be preserved or lost, never recovered, so keep the original and generate everything else from it.