How to batch upscale images at 2x or 4x without uploading them
· 6 min read
To batch upscale images, drop your folder into a batch image upscaler, choose 2x or 4x and run it: every file is enlarged in the same pass and comes back as a PNG, several files together in a zip. AstralBatch does this in your browser with the pica resampling library, so the images are never uploaded.
TL;DR: batch upscaling enlarges every image in a folder by the same factor. AstralBatch resamples the pixels (2x doubles each side, 4x quadruples it) rather than inventing detail with an AI model, so the result is bigger and clean but not more detailed. Output is always PNG, named like photo_2x.png. Start from the best source you have and test one file first.
How do you batch upscale images?
- Open the batch upscale tool and drop in your PNG, JPG or WebP files.
- Choose Upscale 2x or Upscale 4x. The scale applies to the whole run.
- Start the run. Two images are processed at a time, and a progress bar shows how many are done.
- Download the result. One file downloads directly, several arrive as astral_batch_enhanced.zip.
- Open a couple of results at full size before you use the rest.
What does batch upscaling actually do to an image?
It computes new pixels from the ones that already exist. Doubling a 1000 by 750 image gives 2000 by 1500, which means three of every four pixels in the result are new and have to be calculated from their neighbours. AstralBatch uses pica, which resizes with its default mks2013 filter (Magic Kernel Sharp 2013). That filter resizes and sharpens in a single step, so edges stay crisp instead of going soft. What it cannot do is add detail that was never captured.
Is this AI upscaling?
No. A model-based upscaler predicts plausible detail and draws it in, which can look better and is also a guess. A resampling upscaler like this one only interpolates, so nothing untrue is added, and it runs on your own machine without downloading a model. The trade is that a very small or blurry source stays small and blurry, just larger. If you want to understand where that line falls, the guide on what upscaling can and cannot recover covers both approaches.
What image upscaling can and cannot recover: why enlarging an image never recovers lost detail.
Should you pick 2x or 4x?
| Scale | Each side | Pixel count | Use it for |
|---|---|---|---|
| 2x | Doubled | Four times as many pixels | Most jobs: sharper web display, a larger print, a minimum-size requirement |
| 4x | Quadrupled | Sixteen times as many pixels | Small sources that must reach a large size; expect softness and very large files |
A 1000 by 750 photo becomes 2000 by 1500 at 2x and 4000 by 3000 at 4x. Start with 2x and go to 4x only if the size really has to be that big. Each downloaded file carries its scale in its name, for example photo_4x.png, so a mixed set stays easy to tell apart.
Why are the upscaled files so large?
Every result is saved as a PNG, which is lossless. A PNG of an enlarged photograph is much larger than the original JPG because it stores every one of the extra pixels exactly. If size matters afterwards, convert the results to WebP or JPG, or run them through the compressor. Doing that as a second step is better than trading quality away during the upscale.
How to batch convert images (PNG, JPG, WebP): turn the PNG results into WebP or JPG in one pass.
What can go wrong with a large batch?
- Memory is the limit. Browsers cap the size of the canvas they will allocate, and the cap varies by device, so a very large photo at 4x may fail. Try 2x for that file.
- A heavily compressed JPG source has blocky artefacts, and upscaling enlarges them along with everything else. Look for the original first.
- Text and faces show flaws the most, because you know exactly what they should look like.
- Do not upscale twice. Going 2x and then 2x again gives the same size as 4x with more processing and no gain.
- pica on GitHub: high quality image resize in the browser, and its default mks2013 filter
- Wikipedia: Lanczos resampling, the classic interpolation method that resampling upscalers are compared with
Batch upscale images in your browser: 2x or 4x, no account, the whole folder at once.
Frequently asked questions
Can I batch upscale images for free without uploading them?
Yes. AstralBatch upscales in your browser tab with the pica resampling library, so the files are never sent to a server. There is no account to create.
What is the difference between a batch upscaler and an AI upscaler?
A resampling upscaler such as AstralBatch computes new pixels from their neighbours and adds no invented detail. An AI upscaler predicts plausible detail and draws it in, which can look better but is a guess. Resampling is the safer choice when accuracy matters.
Can I mix 2x and 4x in one batch?
Not in a single run. The scale applies to every image you process. You can switch between 2x and 4x before adding more images or run a finished batch again at the other scale.
What format do upscaled images come out in?
Always PNG, saved as name_2x.png or name_4x.png. One image downloads directly and several download as a zip. Convert to WebP or JPG afterwards if you need smaller files.
Does upscaling make a blurry photo sharp?
It makes the image larger and keeps edges clean, but it cannot restore detail that was never captured. A blurry source stays blurry at a bigger size.