Lossy vs lossless: what conversion actually costs you
3 min read · Updated 2026-09-12
The single most common misunderstanding in file conversion is the belief that moving to a higher-quality format raises quality. It does not. A lossless format faithfully preserves whatever you give it — including all the damage a previous lossy step already did. This guide is about what is actually happening, and the handful of cases where the choice really matters.
The two kinds of compression
Lossless compression works the way a ZIP file does: it finds redundancy and encodes it more efficiently, and when you decompress it you get back exactly the bytes you started with. FLAC does this for audio, PNG does it for images, and 7z does it for arbitrary files. Nothing is thrown away.
Lossy compression throws things away on purpose. It uses a model of human perception to decide which parts of the signal you are least likely to notice, and then discards them so the remainder compresses far better. MP3, JPEG, AAC, and H.264 all work this way. The result is dramatically smaller and it is not recoverable.
There is a third category worth naming: uncompressed. WAV and BMP do not compress at all — they store raw samples or raw pixels. Uncompressed is lossless by definition, but it is not the same thing, and it is why WAV files are so much larger than FLAC files that contain identical audio.
The ratchet only turns one way
Once a lossy step has happened, the discarded information is gone. No later conversion can bring it back, because nothing in the file records what was removed.
This means converting an MP3 to WAV produces a large file containing exactly the audio that was in the MP3 — no better, just bigger. Converting a JPG to PNG produces an image with exactly the detail the JPG had, in a format that happens to compress photographs poorly, so the file often grows several times over while looking identical.
Both conversions are sometimes the right thing to do. They are just never quality improvements, and it is worth being clear about which one you are doing.
| Conversion | What happens |
|---|---|
| FLAC → WAV | Same audio, much larger file. Both lossless. |
| FLAC → MP3 | Permanent loss. Small file. Fine for listening. |
| MP3 → WAV | No gain. Large file. Useful only if a tool demands WAV. |
| MP3 → AAC | Second lossy pass. Slightly worse than the MP3. |
| PNG → JPG | Permanent loss, transparency flattened, much smaller. |
| JPG → PNG | No gain. Usually a bigger file. |
| PNG → WebP | Smaller, transparency kept, quality effectively unchanged. |
Generation loss: the case where it really bites
A single lossy conversion from a good source is usually invisible. The problem is repetition. Each time you decode a lossy file and re-encode it, the encoder is compressing artifacts as though they were real detail, and adding its own on top.
Video is where this shows up fastest, because our conversions always re-encode rather than remux. An MP4 converted to MKV, then to AVI, then back to MP4, has been through three full encode passes and will look visibly softer than the original.
The rule that follows is simple: always convert from the best source you have, not from the last thing you converted. Keep the original. Convert once per destination.
When lossless genuinely matters
- You are going to edit the file. Every edit-and-save cycle on a lossy format is another generation. Work in lossless, export to lossy at the end.
- You are archiving a master. If this is the copy everything else will be made from, it needs to hold everything.
- The content has sharp edges. Screenshots, text, line art, and logos develop visible ringing under lossy compression in a way photographs do not. PNG or WebP-lossless, not JPEG.
- You need transparency. Lossy JPEG cannot store it at all.
- The file is evidence, a medical image, or anything where altered pixels are a problem in themselves.
When lossy is obviously right
For anything that is going to be looked at or listened to once and not edited — a photo on a web page, a video someone will stream, a podcast episode, a document scan being emailed — lossy is the correct choice and the size difference is enormous. A well-encoded lossy file at a sensible quality setting is indistinguishable from lossless in normal use, which is the entire point of the design.
The mistake is not using lossy. The mistake is using lossy for the master copy, or converting through it repeatedly, and then being surprised when quality has drained away.