AI Image Upscaler 4x: From $0.01, Priced by Size, No Login

What does AI Image Upscaler do?

AI Image Upscaler 4x makes your images larger (2x, 3x or 4x) with an AI super-resolution model, in bulk. It sharpens edges, cleans JPEG blocks and noise, and returns a new PNG, JPEG or WebP file for each image. Give it image links, upload files from your computer, or paste base64 data from your own code. Each enlarged image is saved in the run's key-value store, and a dataset row gives you its link, the size before and after, and the status.

It runs on CPU, so there is no GPU queue and no third-party AI service. Failed images are free. Use it as an AI image upscaler API: send image links in one request and read the enlarged file links from the dataset.

An alternative to Topaz Gigapixel and Let's Enhance for bulk upscaling

This Actor is not Topaz Gigapixel or Let's Enhance and is not affiliated with either. It enlarges images 2x, 3x or 4x with a Real-ESRGAN model that runs on CPU inside the Actor, up to 5,000 images per run, and you pay per started 640 x 480 block of input, from $0.007 to $0.010 depending on your plan (see the pricing section below).

Why upscale images with this Actor?

• Real 4x. A 512 x 512 image becomes 2048 x 2048. Output goes up to 16 megapixels by default and 64 at most.
• Bulk. Up to 5,000 images per run, from links, uploads or base64, in one dataset.
• Noise control. One setting decides how much grain and JPEG noise is cleaned, from 0 to 100.
• Transparency kept. PNG and WebP output keep transparent areas, so logos and cut-out product shots stay usable.
• Honest billing. You pay only for images that were enlarged and saved.

Typical uses: product photos for a shop, old or small photos, thumbnails for print, game and web art, and images made by AI generators that come out at 1024 px.

How to upscale images, step by step

1. Open the Actor and click Try for free.
2. Paste your image links in Image URLs, or use Upload images to add files from your computer.
3. Pick a Scale: 2x, 3x or 4x.
4. Pick an Output format: PNG (lossless), JPEG or WebP (smaller files).
5. Click Start. When the run ends, open the Enlarged images table and click a link to download each file.

To call it from code, Make, Zapier or n8n, send the same input as JSON through the Apify API and read the dataset. Every row has outputUrl.

How much does it cost to upscale images?

You pay per started 640 x 480 block of input pixels (307,200 pixels, or 0.3 megapixels), not per image. A small image is one block. A larger image is several blocks. The number of blocks is the input width times the input height, divided by 307,200, rounded up.

The price per block depends on your Apify plan: $0.010 on Free, $0.009 on Starter, $0.008 on Scale and $0.007 on Business.

Input size	Blocks	Price on Free	Price on Business
640 x 480	1	$0.010	$0.007
800 x 600	2	$0.020	$0.014
1024 x 768	3	$0.030	$0.021
1280 x 720	3	$0.030	$0.021
1920 x 1080	7	$0.070	$0.049
2048 x 2048 (largest allowed)	14	$0.140	$0.098

Inputs above 4 megapixels, for example 4000 x 3000, are rejected and not charged. Images that fail, and images already at your output limit, are not charged either.

Tip: the price follows the size of the input, so shrinking a large input before you run it costs less.

Worked example: 500 product thumbnails at 400 x 400 (one block each), enlarged 4x to 1600 x 1600, cost 500 x $0.010 = $5.00 on the Free plan ($3.50 on Business). You can set a spending limit on any run, and the Actor stops cleanly when it is reached.

Input example

{
  "imageUrls": ["https://raw.githubusercontent.com/xinntao/Real-ESRGAN/master/inputs/ADE_val_00000114.jpg"],
  "scale": "4",
  "outputFormat": "png",
  "noiseReduction": 50,
  "maxOutputMegapixels": 16
}

Output example

{
  "inputIndex": 1,
  "imageUrl": "https://raw.githubusercontent.com/xinntao/Real-ESRGAN/master/inputs/ADE_val_00000114.jpg",
  "source": "url",
  "status": "ok",
  "outputUrl": "https://api.apify.com/v2/key-value-stores/<store id>/records/upscaled-0001-ADE_val_00000114.png",
  "outputKey": "upscaled-0001-ADE_val_00000114.png",
  "outputFormat": "png",
  "outputWidth": 1024,
  "outputHeight": 1024,
  "inputWidth": 256,
  "inputHeight": 256,
  "scaleRequested": 4,
  "scaleApplied": 4.0,
  "noiseReduction": 50,
  "hasTransparency": false,
  "billedImages": 1,
  "charged": true,
  "error": null
}

Other statuses tell you why an image was skipped, for example not_found, not_image, image_too_large, robots_disallowed or blocked, with a plain-words error.

Tips for the best results

• Start from the largest copy of the image you have. The model adds detail that fits, but it cannot recover text or faces that are only a few pixels tall.
• Use Noise reduction 20 to 40 for clean photos with fine texture, and 60 to 100 for heavily compressed JPEGs.
• Use Maximum output megapixels to hit a size, for example 4 for web pages or 16 for large prints.

Bulk lists

Use a long list of image links from an Apify Dataset or from a file. Pick the Dataset under Dataset with the list and write the name of its field under Field in the dataset. Leave the field name empty to use the first text field. Or paste the web address of a .txt, .csv or .json file under List file URL. A .csv file uses its first column. A header row is skipped. The file can be up to 10 MB. Entries from the Dataset and the file are added to the ones typed under Image URLs, and duplicates are removed. "Maximum images" still limits each run. The price per image does not change.

Related Actors

• Bulk Image Compressor & Converter: Use it to convert, compress and resize images in bulk to WebP, AVIF, JPEG or PNG.
• Image to Text OCR: Use it to extract text from receipts, screenshots, scans and photos.
• Website Screenshot API: Use it to capture full-page screenshots or PDFs of a list of URLs.

FAQ

Is it legal to upscale images with this Actor? Yes for images you own or have the right to edit. The Actor only downloads the links you give it, reads robots.txt on each site first, and sends a user agent with a contact address. It keeps nothing after the run beyond your own storage.

Which model does it use? The open Real-ESRGAN general model (BSD-3-Clause licence), run on CPU. The licence text is in the NOTICE file of the source.

Which file types can I use? PNG, JPEG, WebP, GIF (first frame), BMP and TIFF in; PNG, JPEG or WebP out. Grayscale stays grayscale, and EXIF rotation is applied.

Why did my image fail? The error field says why. The most common causes are a link to a web page instead of the image file, a site that refuses automated downloads, or an input larger than 4 megapixels.

How long does it take? About 2 to 10 seconds per small image with the default 4 GB of memory. Give the run more memory to get more CPU and finish big batches sooner.