AI Image Upscaler API: Upscale & Enhance Photos 2x, 4x
Pricing
from $7.00 / 1,000 image upscaleds
AI Image Upscaler API: Upscale & Enhance Photos 2x, 4x
Upscale images 2x, 3x or 4x with Real-ESRGAN AI: sharpen low-res, blurry or compressed photos, product shots and anime art from image URLs or an Apify dataset. Keeps transparency. $0.01 per image, no usage fees.
Pricing
from $7.00 / 1,000 image upscaleds
Rating
0.0
(0)
Developer
Magnova
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
10 days ago
Last modified
Categories
Security
Limited permissionsShare
AI Image Upscaler API: upscale and enhance photos 2x, 3x or 4x
AI Image Upscaler makes small, blurry or compressed images 2x, 3x or 4x larger with the Real-ESRGAN AI model, adding real-looking detail instead of blur. Give it image URLs or a whole Apify dataset of image links, and get back download links to sharp, high-resolution JPG, PNG or WebP files for $0.01 per image, with no platform usage fees on top.
It is built for batches: product catalogs, scraped listings, AI-generated art, old photos, thumbnails and screenshots. Transparent PNGs keep their transparency, and a dedicated model handles anime, cartoons and illustrations.
What can I use it for?
- E-commerce: upscale supplier or marketplace product photos to 2000x2000 for Shopify, Amazon or print.
- Scraped datasets: point it at the dataset of any scraper run and upscale every
imageUrlin one go. - AI art and print-on-demand: enlarge Midjourney, Stable Diffusion or Nano Banana outputs 4x for posters, T-shirts and stickers.
- Real estate and listings: sharpen low-resolution listing photos.
- Anime, manga and illustrations: clean, crisp line art with the anime model.
- Restoring old or compressed images: remove JPEG artifacts and noise with the denoise slider.
- AI agents and automations: call it from n8n, Make, Zapier or any MCP client with one URL.
How do I upscale an image?
- Paste one or more direct image URLs into Image URLs (or pick a dataset).
- Choose the upscale factor (2, 3 or 4) and the model (photo or anime).
- Click Start. Each result row links to the upscaled file.
Minimal input:
{"images": ["https://upload.wikimedia.org/wikipedia/commons/thumb/3/3a/Cat03.jpg/330px-Cat03.jpg"],"scale": 4}
Upscale every image in a scraper's dataset:
{"dataset": "YOUR_DATASET_ID","imageUrlField": "imageUrl","scale": 2,"outputFormat": "webp"}
The Actor auto-detects common fields (imageUrl, image, images, thumbnailUrl, displayUrl, photo, also one level down such as product.imageUrl). Use imageUrlField with a dot path (images.0, product.mainImage) when your data is shaped differently.
Input options
| Field | What it does | Default |
|---|---|---|
images | Direct image URLs (JPG, PNG, WebP, GIF, BMP, TIFF, AVIF, HEIC) | - |
dataset | Apify dataset whose items contain image URLs | - |
imageUrlField | Field holding the URL in each dataset item | auto |
scale | 2, 3 or 4 times larger per side | 4 |
model | photo (general) or anime (illustrations, cartoons) | photo |
denoise | 0 keeps grain, 1 removes noise and JPEG artifacts strongly | 0.5 |
outputFormat | auto, jpg, png or webp | auto |
quality | JPG/WebP quality 50-100 | 92 |
maxOutputMegapixels | Size limit per result (max 36 MP) | 16 |
Output
Each image becomes one dataset row, and the file itself is saved in the run's key-value store:
{"sourceUrl": "https://upload.wikimedia.org/wikipedia/commons/thumb/3/3a/Cat03.jpg/330px-Cat03.jpg","status": "ok","outputUrl": "https://api.apify.com/v2/key-value-stores/.../records/0001-330px-Cat03-upscaled.jpg","outputFormat": "jpg","inputWidth": 330, "inputHeight": 330,"outputWidth": 1320, "outputHeight": 1320,"scale": 4,"model": "photo","denoise": 0.5,"hasTransparency": false,"outputMegapixels": 1.74,"outputSizeBytes": 316262,"billedExtraMegapixels": 0,"processingSecs": 2.2}
Failed images get status: "error" with a plain-English error (for example "URL returned a web page, not an image") and are not charged. A SUMMARY record in the key-value store lists every image and the counts. The dataset's Overview view shows the original and upscaled images side by side.
How much does it cost to upscale images?
$0.01 per image for results up to 4 megapixels (for example 2000x2000 or 2400x1600). That is all you pay: platform compute is included, so there is no separate usage bill.
- 100 images: $1. 1,000 images: $10.
- Higher Apify plans get volume discounts: Silver $0.0085, Gold and above $0.007 per image.
- Results bigger than 4 MP (you allow them with
maxOutputMegapixels, default 16) add $0.003 per started extra megapixel. Example: a 3200x3200 result (10.2 MP) costs $0.01 + 7 x $0.003 = $0.031. - Unreachable, unreadable, too-large and skipped images are free.
- Set a maximum charge per run and the Actor stops (or shrinks the last result) to stay within it.
Other AI upscalers on the Apify Store charge around $0.02 per image plus platform usage.
How good is the quality?
The photo model is Real-ESRGAN general v3, one of the most used open upscalers (over 100 million runs on public AI APIs). It reconstructs edges, hair, fur, text and textures, and removes JPEG blocks and noise. Because it is not a diffusion model, it stays faithful to your image: it does not invent new faces, objects or text.
The anime model is Real-ESRGAN anime video v3, tuned for flat colors and clean lines.
Large inputs are handled automatically: images above about 500x500 pixels are processed at a reduced working size and rebuilt by the model at the target size, which keeps runs fast and prices flat. The output always has the exact width and height you asked for.
Use with AI agents and MCP
The Actor is ready for AI agents through the Apify MCP server and the Apify API. One field is enough:
{ "images": ["https://example.com/product.jpg"] }
The response row includes outputUrl, the final width and height, and the format, so an agent can hand the link straight to the next step (upload to a store, send by email, attach to a ticket). It runs pay-per-event with limited permissions and works with agentic payments.
Integrations
- n8n / Make / Zapier: run the Actor with the Apify node or module, then read
outputUrlfrom the dataset. - Chain after a scraper: add this Actor as a webhook or integration on your scraper's run and pass its dataset ID.
- API:
POST https://api.apify.com/v2/acts/Magnova~ai-image-upscaler/run-sync-get-dataset-itemswith the JSON input.
FAQ
How do I upscale an image without losing quality?
Use scale 2 or 4 with the photo model and outputFormat png or jpg at quality 92+. Real-ESRGAN adds detail rather than stretching pixels, so edges stay sharp.
Can it upscale images in bulk?
Yes. Paste hundreds of URLs or pick a dataset with thousands of items. Images are downloaded while the previous one is being upscaled.
Does it keep transparent backgrounds?
Yes. PNG, WebP and GIF transparency is preserved; auto format saves those images as PNG.
Which image formats are supported?
JPG, PNG, WebP, GIF (first frame), BMP, TIFF, AVIF and HEIC in; JPG, PNG or WebP out.
What is the biggest image it can make?
Up to 36 megapixels per image (about 6000x6000). The default limit is 16 MP.
Is it a Topaz, Let's Enhance or Upscayl alternative?
It uses the same family of open models as Upscayl, runs in the cloud with no GPU or install, and charges per image instead of a subscription.
Is it legal to upscale images?
Upscaling is processing of files you provide. Only use images you own or have the rights to edit. The Actor does not download from social media pages and does not keep your images beyond your run's storage retention.
Where are my files stored?
In the run's default key-value store on your Apify account. Links in outputUrl work without your token; download them or move them to your own storage, since Apify keeps run storage for your plan's retention period.
Legal and data
Pure computation on images you supply by URL. The models are Real-ESRGAN by Xintao Wang et al., released under the BSD-3-Clause license. No personal data is collected; images are fetched only from the URLs you provide, and results stay in your own Apify storage.