MP3, WAV, FLAC or OGG: Which Format to Use When You Are Editing
Data engineer who loves building high-performance data and web-related tools. Creator of SlowedReverbMaker.net, implementing browser-side digital signal processing (DSP) to democratize audio editing.
Almost every audio format comparison answers the question 'which one sounds best', concludes that lossless is better than lossy, and stops there. It is a correct answer to a question most people are not asking.
If you are editing — slowing a track down, boosting it, adding reverb, cutting it for a video — the question that actually determines your result is different. It is not how good the format sounds. It is how much damage the format does each time your audio passes through it, and how many times that is going to happen before anyone hears it.
Try this while you read: Slowed Reverb Converter
Convert MP3, WAV, FLAC, or OGG and export in either format. Free, in your browser, no signup.
1. The Only Distinction That Matters
There are two families of audio format, and everything else is detail.
- Lossless
- Keeps every sample. WAV stores audio uncompressed as raw numbers; FLAC compresses those numbers the way a ZIP file compresses text, and unpacking returns exactly what went in. A FLAC and a WAV of the same audio are mathematically identical.
- Lossy
- Throws data away permanently. MP3 and OGG Vorbis analyse the audio, decide which parts the ear is least likely to notice, and discard them. Savings are often ten to one, and the discarded detail is gone for good.
- Generation
- One pass through a lossy encoder. Losses accumulate across generations rather than overlapping, because each encoder analyses audio the previous one already damaged.
The Consequence Is Not What People Expect
The important point is not that lossy sounds worse. At 320 kbps most people cannot reliably tell the difference in a single pass.
The important point is what happens when you do it twice — which, if you are editing, you almost certainly are.
2. Generation Loss, and Why It Compounds
Encode a track to MP3 and the encoder makes judgements about what to remove, based on the audio it was handed. Decode it, edit it, and encode it again, and the second encoder makes the same kind of judgements — but now it is analysing audio that has already been damaged, and it cannot restore what the first pass removed. It just makes its own cuts on top.
Count Your Generations
Losses concentrate in exactly the places lossy encoding is weakest: cymbals, reverb tails, sibilance, and anything with a complex, noisy spectrum. Now count the passes in a realistic workflow:
- 1. The track was encoded when it was released or uploaded.
- 2. Encoded again if it came off a streaming or video platform.
- 3. Exporting MP3 from an editor.
- 4. Dropping that into a video editor and rendering.
- 5. Posting the result to a social platform, which re-encodes everything.
3. Why Editing Makes It Worse
Editing is not a neutral operation applied to whatever the format left you. It actively exposes what lossy encoding removed, and slowed edits are the extreme case.
Slowing Stretches the Artefacts Too
A lossy encoder decides what is inaudible by analysing the audio at its original speed, relying heavily on masking — the fact that a loud sound hides a quieter one nearby in time and frequency. Those judgements are only valid at the speed they were made for.
Slow the track down and every artefact stretches with it. A brief smear on a hi-hat that was comfortably masked at full tempo becomes a longer, more exposed sound at 0.85x, where nothing is moving fast enough to hide it any more.
And Reverb Extends Them Further
Reverb is the one effect that deliberately lengthens everything it touches. Artefacts you have already stretched now get a tail attached to them.
This is the real reason a slowed edit made from a bad rip sounds cloudy rather than spacious — and no amount of adjusting the reverb fixes it, because the problem entered the signal before the reverb did.
Boosting Amplifies the Cheapest Data
Encoders allocate the fewest bits to whatever is quietest. So a bass boost amplifies precisely the part of the spectrum that was encoded most crudely.
4. The File Size Maths, Done Properly
Format decisions usually come down to size, so it is worth being able to compute it rather than remembering rules of thumb. Uncompressed audio has a completely predictable size: sample rate × bit depth × channels.
At CD quality — 44,100 samples per second, 16 bits each, two channels — that is 44,100 × 2 bytes × 2, or 176,400 bytes per second.
- WAV — about 10.6 MB per minute. A 3-minute song is roughly 32 MB.
- FLAC — typically 40–60% smaller than the WAV, so roughly 4–6 MB per minute, with identical audio.
- MP3 at 320 kbps — 2.4 MB per minute (320,000 bits per second is 40,000 bytes per second).
- MP3 at 128 kbps — 0.96 MB per minute, and audibly compromised on anything you intend to slow down.
What the Numbers Tell You
The interesting comparison is FLAC against MP3. FLAC gives you a perfect copy for roughly twice the size of a high-bitrate MP3 — not ten times, twice. For archival and for anything you might edit later, that is an easy trade.
The other useful observation is that WAV's size is entirely predictable, which makes it a sanity check. If someone hands you a 3-minute 'WAV' that is 4 MB, it is not 44.1 kHz stereo 16-bit, and something has already happened to it.
5. The Formats, Briefly
With the principles established, the specifics are short.
- WAV — uncompressed, universally supported, large. The right choice as an editing intermediate and whenever anything downstream will encode again.
- FLAC — lossless with roughly half the size of WAV. Ideal for keeping a master copy. Support is good but not universal; some hardware and older software will not open it.
- MP3 — lossy, and supported by literally everything. Its universality is a genuine feature and the reason it has outlived formats that beat it technically. Use 320 kbps if you use it at all.
- OGG Vorbis — lossy, generally better quality than MP3 at the same bitrate, and free of the patent history that shaped MP3's adoption. Its weakness is support: it is not natively playable everywhere, particularly across Apple's ecosystem, which is usually enough to rule it out for sharing.
- AAC / M4A — lossy, better than MP3 at low bitrates, and the default across Apple platforms and most streaming. A reasonable choice when you know the destination.
6. What To Actually Export
The rule that follows from all of this is short enough to remember: export lossy only when the file is the final artefact.
- Uploading the audio directly to a platform, and nothing else will touch it: MP3 at 320 kbps is fine.
- The edit is going into a video editor: export WAV. The editor will encode when it renders, and handing it a WAV means the only lossy step left is that one.
- You might edit it again later: export WAV or FLAC. Re-opening an MP3 to make one more change starts a new generation every time.
- Archiving something you cannot recreate: FLAC. Lossless at half the size of WAV.
- Your source is already a low-bitrate MP3: still export WAV. It will not repair the source, but it stops you adding a generation on top of one that was already weak.
A Note on Browser Tools
Anything that processes audio in the browser, including the tools on this site, decodes your file into raw samples before doing any work. Once decoded, the original format is irrelevant to the processing — a WAV and an MP3 become the same kind of data in memory, and the effect is applied identically to both.
What the format determined was the quality of the samples that decoding produced. The MP3's missing detail is missing before any processing starts. This is worth understanding because it clarifies where quality is won and lost: the source decides the ceiling, the processing decides how close you get to it, and the export format decides how much survives the trip to whoever is listening.
Which is why the slowed reverb converter offers both MP3 and WAV output rather than picking for you. They are answers to different questions.
Method
File size figures are computed from first principles rather than quoted: uncompressed PCM at 44.1 kHz, 16-bit, stereo is 44,100 × 2 × 2 = 176,400 bytes per second, or 10.584 MB per minute; a 320 kbps stream is 40,000 bytes per second, or 2.4 MB per minute. FLAC's compression ratio is content-dependent and the 40–60% range is typical rather than guaranteed — dense, loud material compresses least.