You export a photo as JPEG, send it to a client, and they say the text looks blurry. You go back to your project, the original looks fine. You export again, same result. The problem isn't your settings. It's that JPEG already made its decision the first time, and nothing you do now undoes that.
That's the practical entry point into lossless versus lossy compression. Not definitions, not diagrams. Just: some formats throw away data permanently, and others don't.
The actual difference
When a file is compressed losslessly, the algorithm finds patterns in the data and encodes them more efficiently. Nothing is removed. When you open the file later, you get back exactly what you put in, byte for byte. PNG works this way. So does FLAC for audio. ZIP does too, though the gains are smaller when the content is already compressed.
Lossy compression takes a different approach. It decides what you probably won't notice and discards it. JPEG analyses an image and throws away fine color detail the eye isn't particularly sensitive to. MP3 strips out frequencies your brain masks at normal listening volumes. The result is a smaller file that looks or sounds close enough to the original. Close, but not identical. And what's gone doesn't come back.
This is not a flaw in lossy formats. It's the design. At reasonable quality settings, the difference is genuinely hard to perceive. The problem only shows up when you don't understand which category your format falls into.
Why converting doesn't fix what's already broken
Here's the one thing people get wrong most often: converting a lossy file to a lossless format doesn't improve it.
If you take an MP3 and convert it to WAV, you get a large WAV file. The audio is not better. WAV is lossless, but it can only store what it receives. The frequencies MP3 discarded when encoding the original file are not in the MP3 anymore. The WAV has nothing to restore them from.
Same thing with images. Open a compressed JPEG in Photoshop, save it as PNG. You now have a lossless PNG. The blocky artifacts and smeared edges from the JPEG encoding are still there, preserved faithfully. You've just wrapped degraded data in a lossless container.
The format of a file tells you about its encoding scheme, not the quality of the content inside it.
Re-saving lossy files makes things worse each time
This is called generation loss, and it compounds faster than you'd expect.
Take a JPEG, open it, make a small crop, save it again. The encoder runs again. It makes new decisions about what to discard. Some of what survived the first pass gets thrown out this time. Do it a few more times and you start seeing visible blocking, smeared edges in areas that were originally sharp, color shifts around high-contrast boundaries.
At quality 90%, the degradation from a single re-save is subtle. At quality 70%, it's noticeable within two or three generations. At quality 50%, it's immediate.
The same applies to audio. Re-encoding an MP3 to another MP3 is not neutral. You're running a lossy algorithm on data that was already processed by a lossy algorithm. The artifacts layer.
What quality settings actually mean
For JPEG, the quality slider roughly maps like this: above 85, files are large and visually indistinguishable from lossless at normal viewing distances. Between 70 and 85, files are considerably smaller and still look excellent on screen. Below 60, you start to see blocking in smooth gradients and ringing around sharp edges. Below 50, the artifacts are obvious.
For MP3, 320 kbps is effectively transparent for most audio content and most listeners. 192 kbps is the common sweet spot for streaming, good quality without the file size of 320. 128 kbps is acceptable for voice and simple music, but cymbals and reverb tails start sounding grainy. Below 128, the damage becomes hard to ignore.
These aren't absolute thresholds. Highly detailed images or complex audio will show artifacts sooner. Simpler content holds up better.
The four types of conversions
Not all conversions carry the same risk.
Converting between two lossless formats (PNG to BMP, WAV to FLAC) is genuinely safe. You're just changing the container, the data transfers intact.
Converting from lossless to lossy (PNG to JPEG, WAV to MP3) reduces quality once, at the moment of conversion. If you use a sensible quality setting, this is fine. This is the intended use of lossy formats.
Converting from lossy to lossless (JPEG to PNG, MP3 to WAV) produces a larger file with no quality improvement. Sometimes you need this, for example when a tool requires PNG input but you only have JPEG. Just know you're not recovering anything.
Converting from lossy to lossy (re-encoding a JPEG, converting MP3 to a different bitrate) degrades quality again. Avoid this unless you have no better option.
| Conversion | What happens |
|---|---|
| Lossless to lossless | No loss, safe |
| Lossless to lossy | Quality reduced once |
| Lossy to lossless | Larger file, artifacts preserved |
| Lossy to lossy | Quality reduced again |
A workflow that holds up over time
Keep masters in lossless formats. WAV for audio you recorded or produced. PNG for graphics and screenshots. RAW or TIFF for photographs if the work is important. Edit and revise in these formats.
When you need to deliver, export to lossy from the master. MP3 at 192 kbps or above for audio. JPEG at 80 to 85 for most photography. WebP for web images where you want the smallest file with good visual quality.
Never go back and re-export from the lossy delivery file. If the client asks for a different format or the specs change, go back to the master and export again.
The quality loss from lossy compression happens once and stays predictable when you work this way. It only becomes a problem when the master is already lossy, or when you keep re-encoding the same file.
Most of the confusion around these formats comes from not knowing where on this chain a particular file sits. Once that's clear, the decisions mostly make themselves.