Upscale Image
Enlarge images 2x or 4x while keeping sharpness.
Select an image
or drag and drop it here
JPG, PNG, WEBP, GIF, AVIF. Up to 50 MB.
2x & 4x upscale
Up to 8192px
Lanczos3 resampling
Sharp results
No sign-up
Free forever
Files never stored
100% private
Related Tools
Image upscaler
Enlarge images 2x or 4x, free
Sometimes a photo is fine except that it is too small — a marketplace wants a bigger file, a print needs more pixels, a thumbnail spec will not accept what you have. Upscaling rebuilds the image at two or four times the size, keeping it as clean as an enlargement can be.
Upload a JPG, PNG, WEBP, GIF or AVIF up to 50 MB, choose 2x or 4x, and download the result. The exact output dimensions are shown before you run it, so there is no guessing at what you will get.
Below: how the enlargement is actually done and why the method matters, what upscaling can and cannot deliver, and which formats come back as what.
The basics
How the enlargement is done
Enlarging an image means building a bigger grid of pixels and deciding what colour each new one should be. Every method answers that question differently, and the answer is what separates a clean enlargement from a blocky or mushy one.
This tool uses Lanczos3, the sharpest of the classical resampling kernels, followed by a sharpen pass calibrated to the factor you picked — stronger at 4x, where more softening happens. Output is capped at 8192 px on the long edge, and if 4x would exceed that, both sides are scaled down together so the shape is preserved instead of the image being cropped.
What none of that does is add information. The enlargement is built entirely from the pixels the original already had — which is the single most important thing to understand before you upload something small and hope for a miracle.
Methods
Why the resampling method matters
Four ways to fill in a bigger grid, from crudest to sharpest. Most software defaults to the third; this uses the fourth.
Nearest neighbour
Copies the closest existing pixel. Instant, and produces hard blocky squares when enlarging. Its one real use is pixel art, where the blockiness is the point.
Bilinear
Averages the four surrounding pixels. Smoother than nearest neighbour and noticeably soft at any meaningful enlargement.
Bicubic
Looks at a wider neighbourhood and fits a curve through it. The default in most editing software, and a clear step up from bilinear — still soft at 4x.
Lanczos3
Weighs an even wider window with a sinc-based function, which preserves edges and fine texture better than bicubic at the cost of some processing time. This is what runs here, followed by a sharpen pass to counter the softening that any enlargement introduces.
Expectations
What upscaling is good for — and what it is not
The honest version: this solves a size problem, not a quality problem. Knowing which one you have saves a lot of disappointment.
Good for
- A photo that is simply too small for where it needs to go — a marketplace minimum, a thumbnail spec, a print size.
- Clean line art, logos and flat illustrations, where Lanczos3 holds edges very well.
- Screenshots and diagrams that need to display larger without going blocky.
- A modest jump — 2x from a decent original is where this looks best.
- Print, where the constraint is genuinely pixel count rather than detail.
- Scans that came off the scanner at too low a DPI to use.
Not what it does
- Recovering detail. Enlarging spreads the information you have over more pixels; it does not add any.
- Fixing a blurry or out-of-focus shot. There is no edge in the file for the sharpen pass to work with.
- Rescuing a heavily compressed image — 4x enlarges the JPEG artefacts along with everything else.
- Reading text that is already illegible. If you cannot read it at 1x, 4x will not help.
- Turning a thumbnail into a poster. Expectations should scale with how much real information the original holds.
How to use it
Three steps
- 1
Upload
One image at a time — JPG, PNG, WEBP, GIF or AVIF, up to 50 MB.
- 2
Pick 2x or 4x
The exact output dimensions are shown before you run it, so you can check the result will be the size you actually need.
- 3
Download
The enlarged image comes back with no watermark, and you are shown the before and after dimensions.
Formats
What you upload, and what comes back
There is no output-format picker — the result follows the input, with two exceptions worth knowing about before you upload.
| Upload | You get back | Worth knowing |
|---|---|---|
| JPG | JPG | Straightforward — no transparency involved either way |
| PNG | PNG | Transparency is preserved |
| WEBP | WEBP | Transparency is preserved |
| GIF | JPG | A single frame, flattened onto white. Animation is not kept |
| AVIF | JPG | Flattened onto white, so any transparency is lost |
GIF and AVIF come back as JPG. Both are flattened onto white, so transparency is lost and an animated GIF becomes a single frame. If that matters, convert to PNG first and upscale that.
Telling them apart
Upscale, resize, sharpen or enhance?
Four tools that sound similar and solve different problems. Picking the wrong one is the most common reason a result disappoints.
| What it does | Adds pixels? | Reach for it when | |
|---|---|---|---|
| Upscaling | Enlarges the pixel grid with Lanczos3, then sharpens | Yes | Making an image physically bigger for print or a large display |
| Resizing | Scales the pixel grid to any target dimensions, up or down | Yes, mechanically | Hitting an exact width and height |
| Sharpening | Raises contrast across edges that already exist | No | Making a soft image look crisper at its current size |
| Enhancing | Noise, sharpening, clarity, exposure and colour together | Optionally | A photo with several things wrong at once |
Why this one
What you get
Lanczos3, not bicubic
The sharpest classical kernel rather than the default most software ships with, followed by a sharpen pass tuned to the factor you chose.
See the size first
The exact output dimensions appear before you run it, so you know whether 2x clears the requirement or whether you need 4x.
Aspect ratio held
Past the 8192 px ceiling both sides scale down together. An earlier version clamped them independently and quietly cropped wide images; it does not any more.
2x and 4x, both free
No paid tier for the larger factor and no daily cap on either.
Files are never stored
Your image is processed in memory and handed back in the same request. Nothing is written to disk or saved to a database.
No watermark, no sign-up
No account, no email, and a clean file at the end.
Privacy & security
What happens to your image
Your image is uploaded over an encrypted HTTPS connection, decoded, resampled in memory, and handed straight back in the same request. It is never written to disk, never saved to a database and never queued for later. Once the response is sent, nothing about it remains.
No account is required, so an upscale is not attached to an identity. Full detail is in the privacy policy.
Troubleshooting
When the result is not what you expected
The result is softer than I expected
Any enlargement softens — there is no new detail to fill the extra pixels with. A sharpen pass runs afterwards to counter it, stronger at 4x than 2x, but it can only work with edges the original actually has.
My GIF came back as a JPG
GIF has no case in the output switch, so it falls through to JPG, flattened onto white. Convert it to PNG first if you need transparency or a lossless result.
My transparent AVIF lost its transparency
Same reason — AVIF output is not offered, so it comes back as JPG with the transparent areas filled white. Convert to PNG before upscaling.
The output is smaller than 4x
Output is capped at 8192 px on the long edge. Both sides are scaled down together to stay under it, so the aspect ratio is preserved rather than the image being cropped.
The file got much bigger
Four times the pixels is roughly four times the data. A 4x PNG in particular can grow dramatically — compress the result if you need it smaller.
Artefacts look worse after upscaling
Because they were enlarged too. Compression blocking and noise scale up with everything else. Start from the cleanest original you have.
Best practice
Getting the best enlargement
- Start from the largest, least-compressed original you have, even if it is still smaller than you need.
- Prefer 2x. It is the more reliable result, and two things asking for detail rarely both get it.
- Upscale before adding text or graphics, never after — enlarging a text overlay softens it.
- For a transparent source, convert GIF or AVIF to PNG first so the output keeps its alpha.
- Check the predicted dimensions on screen before running it, especially near the 8192 px ceiling.
- Keep the original. You may want a different factor later, and re-upscaling an upscale compounds the softness.
Enlarging a logo or a flat graphic? Tracing it to SVG beats any amount of upscaling — a vector scales to any size with no loss whatsoever.
FAQ
Frequently asked questions
What is image upscaling?
Increasing an image's pixel dimensions so it is physically larger. The pixel grid is rebuilt at the new size by interpolating between the pixels that already exist.
How does this upscaler work?
It resamples the image with Lanczos3 — the sharpest of the classical interpolation kernels — and then runs a sharpen pass to counter the softening any enlargement introduces.
Can upscaling add detail that was not in the original?
No. Enlarging spreads the information the file already holds across more pixels. A good kernel makes that enlargement clean and sharp rather than blocky, but the amount of real detail does not change.
Is Lanczos3 better than the bicubic my editor uses?
For enlarging, generally yes. Lanczos3 weighs a wider window and holds edges and fine texture noticeably better than bicubic, which is why it is used here rather than the more common default.
Should I choose 2x or 4x?
2x if it gets you to the size you need. Every doubling asks the interpolation to fill in more, so 2x from a decent original consistently looks better than 4x from a small one.
Can I upscale a blurry photo?
You can enlarge it, but it will still be blurry — larger. Sharpening works on edges that exist in the file, and blur destroyed them. For a slightly soft photo the photo enhancer's Sharpen mode is the better tool.
Which formats can I upload?
JPG, PNG, WEBP, GIF and AVIF.
What format do I get back?
The same format for JPG, PNG and WEBP. GIF and AVIF come back as JPG, flattened onto a white background — so convert those to PNG first if you need transparency.
Is transparency preserved?
For PNG and WEBP, yes. For a transparent AVIF or GIF it is not, because the output falls back to JPG, which has no alpha channel.
What happens to an animated GIF?
Only one frame survives. The output is a still JPG — this tool has no animated output format.
Is there a limit on the output size?
Yes, 8192 pixels on the long edge. If 4x would exceed it, both sides are scaled down together so the aspect ratio is kept intact rather than the image being cropped.
What is the maximum upload size?
50 MB per image.
How many images can I upscale at once?
One at a time.
What is the difference between upscaling and resizing?
Upscaling is enlargement specifically, tuned for it — Lanczos3 plus a sharpen pass, at fixed 2x or 4x factors. The resizer scales to any dimensions you type, up or down. If you need an exact width and height rather than a multiple, use that instead.
What is the difference between upscaling and enhancing?
Upscaling changes the size. Enhancing changes the quality — noise, sharpness, exposure and colour. The photo enhancer includes an Upscale mode that does both in one pass if you need it.
Can I upscale artwork, anime or illustrations?
Yes, and this is where Lanczos3 does its best work. Clean lines and flat colour enlarge far better than photographic texture does.
Can I upscale a logo?
Yes, though if the logo is a flat graphic, tracing it to SVG is the better answer — a vector scales to any size with no loss at all.
Can I upscale product photos for a marketplace?
Yes, and meeting a minimum resolution requirement is one of the most common reasons people use this. Start from the largest original the photographer gave you.
Will upscaling help before printing?
It helps where the problem is genuinely pixel count — a print needs far more pixels per inch than a screen. It does not make a low-detail photo look detailed at poster size.
Why did my file get so much bigger?
A 4x upscale has sixteen times as many pixels as the original. PNG output in particular grows sharply. Run the result through the compressor if the size matters.
Does it add a watermark?
No. The upscaled image downloads clean.
Is it free?
Yes, both factors, with no account, email or sign-up.
Do I need to install anything?
No. You upload from this page and download the result.
Are my images stored?
No. Your image is processed in memory and returned in the same request. Nothing is written to disk or saved to a database.
Can I use it on mobile?
Yes, it works in mobile browsers on Android and iPhone.
Something not covered? There are more answers on the FAQ page, or you can contact support.
Related tools