🔍 AI Image Upscaler Scraper

🚀 Export sharper, bigger images in seconds. Paste image links, get 2x or 4x Real-ESRGAN upscales with public download links: a 500x500 photo becomes 2000x2000 in about 3.5 seconds.

The AI Image Upscaler takes public image URLs, enlarges each image 2x or 4x with the Real-ESRGAN general model, and saves the result in the run's key-value store with a public link. Every row also reports the source and output dimensions, formats, file sizes and the processing time, so you can audit a whole batch from the dataset.

Measured on the Apify platform: a 500x500 JPEG turns into a 2000x2000 PNG in 3.5 s, and a 1-megapixel photo into a 4096x4096 (16.8 MP) PNG in 19.4 s. Each row carries 19 fields. Up to 500 images per run, transparency is kept, and URLs that are not images are skipped without a charge.

🎯 Target Audience	💡 Primary Use Cases
E-commerce teams, marketers, designers, print shops, archivists, developers and AI agents	Product photo enlargement, print-ready scans, thumbnail rescue, restoring old or compressed photos, batch pipelines in Make, Zapier or n8n

📋 What the AI Image Upscaler Scraper does

• 🔍 Upscales 2x or 4x with Real-ESRGAN (general x4v3), a model trained to rebuild edges and texture and to clean JPEG artifacts, not just stretch pixels.
• 🔗 Returns public links to every upscaled file, ready to download, embed or pass to the next step of a workflow.
• 🗂 PNG, JPEG or WebP output, with adjustable quality for JPEG and WebP.
• 🫥 Keeps transparency in PNG and WebP: the alpha channel is enlarged and re-attached.
• 📏 Reports the numbers: source and output width, height, format and bytes, megapixels and seconds per image.
• 🛡 Fails safely: web pages, PDFs, videos, broken links and private-network addresses become error rows that are never billed.

💡 Why it matters: upscaling services usually mean a subscription, credits that expire or a desktop app. Here you send a list of links through the API and get files back in one dataset you can export to CSV, Excel, JSON or XML.

🎬 Full Demo

🚧 Coming soon

📊 Output

Field	Description
🖼 imageUrl	Public link to the upscaled file
🔗 sourceUrl	The image URL you provided
📄 fileName	Key of the file in the run's key-value store
🔍 scale	Upscale factor applied (2 or 4)
↔️ outputWidth / ↕️ outputHeight	Size of the result in pixels
🧮 outputMegapixels	Result size in megapixels
🗂 outputFormat	png, jpeg or webp
💾 outputSizeBytes	Size of the result file
↔️ inputWidth / ↕️ inputHeight	Size of the source image (after EXIF rotation)
🗂 inputFormat	Detected source format
💾 inputSizeBytes	Size of the downloaded source file
📉 inputDownscaled	Yes when a source above 1 MP was shrunk to 1 MP first
🫥 hasTransparency	Yes when the source had transparent pixels
🧠 model	Upscaling model used
⏱ processingTimeSec	Seconds spent upscaling and encoding
🕒 scrapedAt	When the image was processed
❌ error	Why an image was skipped (null on success)

Three real rows from a platform run:

{
  "imageUrl": "https://api.apify.com/v2/key-value-stores/iyepbKMNwSfLERsSc/records/upscaled-001-mugbe3ws.png?signature=1cLe2ToKfJlpe2rXkwrNF",
  "sourceUrl": "https://upload.wikimedia.org/wikipedia/commons/thumb/9/97/The_Earth_seen_from_Apollo_17.jpg/500px-The_Earth_seen_from_Apollo_17.jpg",
  "fileName": "upscaled-001-mugbe3ws.png",
  "scale": 4,
  "outputWidth": 2000,
  "outputHeight": 2000,
  "outputMegapixels": 4,
  "outputFormat": "png",
  "outputSizeBytes": 6175073,
  "inputWidth": 500,
  "inputHeight": 500,
  "inputFormat": "jpeg",
  "inputSizeBytes": 82862,
  "inputDownscaled": "No",
  "hasTransparency": "No",
  "model": "Real-ESRGAN general x4v3",
  "processingTimeSec": 3.5,
  "scrapedAt": "2026-09-25T02:01:13.373Z",
  "error": null
}

{
  "imageUrl": "https://api.apify.com/v2/key-value-stores/iyepbKMNwSfLERsSc/records/upscaled-003-mugbemji.png?signature=1SJ8wo5No24v7QRqqRjki",
  "sourceUrl": "https://upload.wikimedia.org/wikipedia/commons/thumb/9/97/The_Earth_seen_from_Apollo_17.jpg/1280px-The_Earth_seen_from_Apollo_17.jpg",
  "fileName": "upscaled-003-mugbemji.png",
  "scale": 4,
  "outputWidth": 4092,
  "outputHeight": 4096,
  "outputMegapixels": 16.76,
  "outputFormat": "png",
  "outputSizeBytes": 24554250,
  "inputWidth": 1280,
  "inputHeight": 1281,
  "inputFormat": "jpeg",
  "inputSizeBytes": 431044,
  "inputDownscaled": "Yes",
  "hasTransparency": "No",
  "model": "Real-ESRGAN general x4v3",
  "processingTimeSec": 19.4,
  "scrapedAt": "2026-09-25T02:01:38.032Z",
  "error": null
}

{
  "imageUrl": "https://api.apify.com/v2/key-value-stores/iyepbKMNwSfLERsSc/records/upscaled-005-mugbf5sn.png?signature=N3CIfu5fOS2H1ICiBNwN",
  "sourceUrl": "https://upload.wikimedia.org/wikipedia/commons/thumb/2/27/Apollo_11_insignia.png/330px-Apollo_11_insignia.png",
  "fileName": "upscaled-005-mugbf5sn.png",
  "scale": 4,
  "outputWidth": 1320,
  "outputHeight": 1332,
  "outputMegapixels": 1.76,
  "outputFormat": "png",
  "outputSizeBytes": 2275587,
  "inputWidth": 330,
  "inputHeight": 333,
  "inputFormat": "png",
  "inputSizeBytes": 140339,
  "inputDownscaled": "No",
  "hasTransparency": "Yes",
  "model": "Real-ESRGAN general x4v3",
  "processingTimeSec": 2.1,
  "scrapedAt": "2026-09-25T02:02:02.277Z",
  "error": null
}

✨ Why choose this Actor

• ⚡ Fast on ordinary images: 2 to 4 seconds for a typical web image, about 20 seconds for a full 1-megapixel photo at 4x.
• 📦 Batches, not one file at a time: send up to 500 links in a single run and get one dataset back.
• 🔢 Real output sizes: 4x means 4x. A 1 MP photo comes back at about 16 MP, far beyond the 1-2 MP targets of diffusion upscalers.
• 🧾 Transparent about every file: dimensions, bytes and timing per row, so you can check a batch without opening each image.
• 🔒 No external AI service: images are processed inside the run and stored in your own Apify storage.
• 🆓 Failed images are free: only successfully stored results count.

📈 How it compares to alternatives

	This Actor	Desktop upscalers	Subscription web tools
Batch of links in one call	✅ Up to 500 per run	❌ Manual file picking	⚠️ Usually a few at a time
API, Make, Zapier, n8n	✅ Native	❌	⚠️ Paid tiers only
Output size	2x or 4x of the source (source capped at 1 MP)	Up to 6x	Often capped by plan
Transparency kept	✅ PNG and WebP	✅	⚠️ Varies
Per-image metadata	✅ 19 fields	❌	❌
Face-specific restoration	❌ Not included	✅ Some	✅ Some

Honest ceiling: sources larger than 1 megapixel are shrunk to 1 megapixel before the 4x model runs (the row says so in inputDownscaled), so the largest output is about 16 megapixels. There is no dedicated face restoration model.

🚀 How to use

1. Create a free Apify account with $5 in monthly credit: sign up here.
2. Open the AI Image Upscaler and paste your image links into Image URLs (one per line, or upload a text file of links).
3. Pick 4x or 2x and the file format (PNG, JPEG or WebP).
4. Click Start. Each image takes a few seconds.
5. Open the Output tab: click any imageUrl to download the file, or export the dataset to CSV, Excel, JSON or XML.

Example run settings:

{ "imageUrls": [{ "url": "https://upload.wikimedia.org/wikipedia/commons/thumb/9/97/The_Earth_seen_from_Apollo_17.jpg/330px-The_Earth_seen_from_Apollo_17.jpg" }], "scale": "4", "outputFormat": "png" }

💼 Business use cases

🛒 E-commerce and marketplaces

Enlarge supplier product shots that arrive at 400 or 500 pixels so they pass marketplace minimum-size rules and zoom cleanly on product pages.

📣 Marketing and social media

Turn small logos, screenshots and archive photos into banner-ready assets without hunting for the original files.

🖨 Print and publishing

Push web-resolution photos toward print sizes: a 500 px image at 4x reaches 2000 px, enough for a small print at 300 dpi.

🗄 Archives and real estate

Batch-upscale scanned photos or low-resolution listing images, keeping a dataset that records exactly what was processed and when.

🔌 Automating AI Image Upscaler Scraper

• Make and Zapier: trigger a run when a new image lands in a folder or a spreadsheet row, then send the imageUrl onward.
• n8n: call the Actor from an HTTP node and loop over the dataset rows.
• Slack: post the finished links to a channel with the Apify Slack integration.
• Google Drive: copy the upscaled files into a shared folder after each run.
• GitHub Actions: upscale documentation screenshots as part of a build.
• Airbyte: sync the run metadata into your warehouse for reporting.

🌟 Beyond business use cases

• Research: prepare enlarged figures and satellite images for papers and slides.
• Personal: give old family photos a second life before printing them.
• Non-profit: restore small archive images for exhibitions and websites.
• Experimentation: compare 2x and 4x results or feed upscaled frames into other AI tools.

🤖 Ask an AI assistant about this scraper

Copy this Actor's URL into ChatGPT, Claude or Perplexity and ask how to connect it to your workflow, which scale to use for a print size, or how to loop over the dataset in your language of choice.

❓ Frequently Asked Questions

🖼 Which image formats can I send?

JPEG, PNG, WebP, GIF (first frame), TIFF and AVIF, up to 25 MB per file. iPhone HEIC/HEIF photos are not supported; convert them to JPEG first.

🔗 What kind of link works?

A link that downloads the image file itself. A web page that merely shows the image (a gallery page, a Google Drive share page) is rejected with an error row.

🔍 What is the difference between 2x and 4x?

4x multiplies width and height by four (16 times the pixels). 2x is the 4x result resized down to twice the original size, which keeps the restored detail and gives smaller files.

📏 Is there a size limit?

Sources above 1 megapixel are shrunk to 1 megapixel first, so a 4x result is at most about 16 megapixels. Files above 25 MB or 50 megapixels are refused.

⏱ How long does it take?

About 3.5 seconds for a 500x500 image and about 20 seconds for a 1 megapixel photo at 4x, measured on the platform with default settings.

🫥 Is transparency preserved?

Yes, in PNG and WebP output. JPEG has no transparency, so transparent areas turn white.

🙂 Does it restore faces?

Faces are upscaled like the rest of the picture. There is no separate face restoration model.

💸 Am I charged for images that fail?

No. Broken links, non-image files and refused URLs produce an error row that is never billed.

🔒 Where are my images stored?

In the run's own key-value store in your Apify account. The public links carry a signature and follow your account's storage retention.

📦 How many images can one run handle?

Up to 500 images per run. Split larger batches into several runs, which can also run in parallel.

🧠 Which model is used?

Real-ESRGAN general x4v3, an open model released under the BSD-3-Clause licence, running inside the Actor.

Credit: the upscaling model is Real-ESRGAN realesr-general-x4v3, Copyright (c) 2021 Xintao Wang et al., used under the BSD-3-Clause licence. The full licence text ships with the model inside the Actor (models/LICENSE-Real-ESRGAN).

🏷 Can I use the results commercially?

The model licence allows commercial use. Rights to the pictures themselves stay with whoever owns the original images.

🔌 Integrate with any app

Use the Apify API, the JavaScript or Python client, webhooks, or ready-made integrations for Make, Zapier, n8n, Slack, Google Drive, Airbyte and more. Every run produces a dataset and a key-value store you can read from any tool.

🔗 Recommended Actors

• Google Lens Scraper: find where an image appears online.
• Public Domain Vectors Scraper: collect free CC0 clipart and vectors.
• Freepik Coloring Pages Scraper: gather printable coloring pages.
• Civitai Models Scraper: track AI image models and creators.

💡 Pro Tip: browse the complete ParseForge collection for more data tools.

🆘 Need Help? Open our contact form

⚠️ Disclaimer: this is an independent tool, not affiliated with the Real-ESRGAN authors or any image host. It only processes publicly available data at the links you provide; make sure you have the right to use and modify those images.