Make a PNG Smaller Without Losing Quality

PNG's compression already reconstructs every original pixel exactly, so making a PNG smaller without losing quality means storing those same pixels more efficiently: better filter choices, a stronger DEFLATE pass, and stripping metadata chunks the image doesn't need. That typically saves 5–25%. The much larger savings — the ones that take 4 MB to 400 KB — come from reducing the number of colours, and that does throw information away, however often it is sold otherwise.

On Android, Convertlix's PNG Optimizer makes PNGs smaller on the handset, so the image never leaves the device — no upload, no server-side copy, nothing to retain. The scope, up front: Android 8.0 or later, sign-in required (there is no guest mode), 25 MB per file and batches of 5 on the free tier against a server-set monthly allowance of conversions shown in the app, and the result lands in Downloads. Everything below applies whichever optimiser you use.

What's actually in your PNG

PNG stores pixels exactly. Every byte you get back has to come from one of four places, and knowing which is which is the whole skill here.

1. Metadata chunks (genuinely free). Colour profiles, text comments, creation timestamps, editor signatures. On a screenshot these are usually a few kilobytes; on an export from a design tool with an embedded ICC profile they can be tens of kilobytes. Removing them changes no pixel. The one caveat: dropping a colour profile can shift how colours are displayed on wide-gamut screens, so keep it if colour accuracy matters.

2. Filter and compression choices (genuinely free). Before compressing, PNG applies a per-row filter that predicts each pixel from its neighbours, and the encoder picks which filter to use. Most software picks quickly rather than optimally. Trying all of them and running a heavier DEFLATE pass produces a smaller file containing byte-identical pixels — this is what tools like oxipng and zopflipng do, and it typically returns 5–25%.

3. Bit depth and palette (not free, often invisible). A 24-bit PNG stores three bytes per pixel whether or not the image uses three bytes' worth of colour. A screenshot of an app interface might contain 200 distinct colours; converting it to an 8-bit indexed PNG stores one byte per pixel and can cut the file by 60–70% with no visible change at all, because nothing was thrown away — every colour present is still present. This is the big win on UI screenshots, logos, diagrams and flat illustrations.

4. Colour quantisation (definitely lossy). If the image has more than 256 colours — a photograph, a gradient, a soft shadow — reducing it to a palette means discarding colours and approximating them. Tools like pngquant do this well, and it is where the dramatic 90% reductions come from, but it is lossy compression with a nicer name. Gradients band, soft shadows go blotchy, and anti-aliased text edges develop a halo.

The most common mistake: it shouldn't be a PNG

Before optimising, check whether the file should exist in this format. PNG is designed for images with flat colour, hard edges and transparency. It is a poor fit for photographs, and the size difference is not marginal.

A 12-megapixel photograph saved as a PNG can easily be 15–25 MB. The same photograph as a good-quality JPEG is 2–4 MB, and at normal viewing size you will not see a difference. PNG has no lossy mode to fall back on, so it stores every sensor-noise speckle faithfully — and noise is the most expensive thing you can ask a compressor to store.

The test is simple:

• Photograph, or anything with continuous tone? JPEG (or WEBP). If you need it under a specific cap, Resize a Photo to Under 100 KB for a Form covers the lever order for that.
• Screenshot, logo, diagram, chart, pixel art, or anything needing transparency? PNG is correct — optimise it rather than converting it.
• Screenshot of a photograph? It is a photograph. Save it as a JPEG.
• A phone screenshot of a text-heavy page? PNG, and palette reduction will probably halve it with no visible change.

WEBP deserves a mention here because it handles both cases: in its pixel-exact mode WEBP is usually 20–30% smaller than an optimised PNG, and lossy WEBP with an alpha channel does something PNG cannot do at all. As of September 2026 browser and OS support is broad, but some older tools, some upload forms and some print workflows still refuse it — so check what the destination accepts before committing.

How to make a PNG smaller

1. Decide whether it should be a PNG at all. Continuous tone means JPEG or WEBP, and no optimiser closes the gap between a PNG photograph and a JPEG one.
2. Check the pixel dimensions first. A 3,000 px-wide screenshot displayed at 800 px is carrying nine times the pixels it needs. Resizing is multiplicative and beats every compression setting — do it before you optimise, not after.
3. Run the pixel-exact pass. In Convertlix, open PNG Optimizer and pick the file through Android's picker; it re-encodes the PNG on the phone. Expect a modest, free saving — this step cannot make the image worse.
4. Count your colours before reaching for a palette. If the image genuinely uses fewer than 256 colours, indexing it is free. If it has gradients or a photograph in it, palette reduction is lossy and you should decide that deliberately.
5. Compare at 100% zoom, on the busiest area. Look at gradients, soft shadows and anti-aliased text edges — those are where quantisation shows first, and they are invisible in a thumbnail.
6. Keep the original. Optimisation writes a new file. If the palette version turns out to band on a projector, you want the 24-bit one back.

When this doesn't work

• The PNG is a photograph. Optimisation returns single-digit percentages on continuous-tone images because there is no redundancy for DEFLATE to find. The fix is a different format, not a better optimiser.
• You need an exact colour match. Print work and brand assets sometimes require a specific ICC profile and full bit depth. Stripping the profile or indexing the palette can shift colours measurably, even when the pixels look identical on your screen.
• The image is already optimised. A PNG exported by a modern build pipeline or a design tool with optimisation enabled has had the free savings taken. Re-running an optimiser on it returns almost nothing, and that is a correct result rather than a failure.
• Very large canvases. A 10,000 × 10,000 px PNG can exceed the 25 MB free-tier file cap before you get to optimise it. That is a limit, not a bug — resize first, or use a desktop tool.
• Animated content. APNG and GIF are different problems with different tools, and a still-PNG optimiser doesn't apply.

FAQ

Does PNG compression throw any detail away?

No — PNG's DEFLATE compression reconstructs the original pixels exactly, which is why it is used for screenshots, logos and anything with hard edges. That also means the pixel-exact savings available to an optimiser are limited: rewriting the filters and running a heavier compression pass typically returns 5–25%. The dramatic reductions advertised by some tools come from reducing the colour palette, which discards colour information and is a lossy operation despite how it is usually described.

Why is my PNG bigger than the JPEG version of the same image?

Because PNG stores every pixel exactly, including sensor noise, while JPEG discards detail your eye is unlikely to miss. On a photograph that difference is enormous — the same 12-megapixel image can be 20 MB as a PNG and 3 MB as a good-quality JPEG. PNG is the right format for flat colour, hard edges and transparency, and the wrong one for photographs.

Will optimising a PNG make my screenshot blurry?

A pixel-exact optimisation cannot: it produces byte-identical pixels in a smaller file. Reducing the colour palette can, if the screenshot contains gradients or an embedded photograph — you would see banding in the gradient rather than blur. A screenshot of flat interface elements usually has few enough distinct colours that indexing it is genuinely free.

Does Convertlix upload my images?

No. The PNG is opened through Android's own file picker, re-encoded on the phone, and saved back to the device — nothing is sent anywhere, and the app has no storage permission, so it cannot read anything you did not hand it. Its network calls carry identity, licence, quota, configuration and payment data, never an image. Sign in once, switch on airplane mode, and the optimiser still works.

Is there a free way to optimise PNGs without an app?

Yes, and the command-line tools are the best in class. oxipng -o 4 file.png does a thorough pixel-exact pass, zopflipng squeezes a little further at the cost of a lot of CPU time, and pngquant --quality=65-90 does the lossy palette reduction well and tells you what it did. Squoosh does all of this in the browser with a live preview of the output size, which makes hitting a specific cap easy. All of these run locally and handle whole folders, which beats any phone batch limit — the reason to use a phone tool is that the screenshot is on the phone.

---

If the PNG is on your phone and something is refusing it for being too big, take the free pixel-exact saving first and only reach for a palette if you still need the bytes.