AI Image Upscaler

Take a small or low-resolution image and get back a 2× or 4× version with the detail intact. Unlike traditional resampling (which just stretches existing pixels and goes blurry), our AI upscaler infers what the missing detail should look like — sharp edges, recognisable textures, faces with actual features. Drop the image, click Upscale, download the result. The first one's free.

How AI upscaling differs from "make it bigger"

If you take a 500-pixel-wide image and stretch it to 2000 pixels in Photoshop or any traditional tool, you get a blurry 2000px image. The information density is identical to the original — there are just bigger pixels. Our upscaler does something fundamentally different: it uses a neural network trained on millions of high-resolution photos to predict what the high-res version of your image would look like. The output has detail that wasn't there in the input, inferred from the model's understanding of how real-world subjects look.

This means upscaling works best on natural images with clear subjects — photos of people, animals, products, landscapes, vehicles. It works decently on illustrations and digital art (use our Anime Enhancer for line-art content specifically). It works less well on heavily-compressed JPGs where the original already has visible artifacts, because the AI tries to upscale the artifacts too. Run them through the Photo Enhancer first to clean them up.

How to upscale an image

1. Drop your image. JPG, PNG, WebP, HEIC all work. Input limits: 10 MP free, 25 MP signed in, 50 MP Premium.
2. Pick a scale factor. 2× doubles each dimension (4× area). 4× quadruples each dimension (16× area). 4× takes about 2× as long to process.
3. Wait 8–15 seconds for the AI to process. Output max is 200 MP — a 4× upscale of an 8 MP source lands at 128 MP.
4. Download the upscaled JPG. Output filename gets a `_2x` or `_4x` suffix.

Common use cases

Print-size enlargement. Got a 1500×1000 photo and want to print it at 20×30 inches? Run a 4× upscale to 6000×4000, then send to print at 200 DPI. The output handles framed prints up to 24 inches per side without visible quality loss.

Old digital photos. Camera tech in 2008 took 5 MP photos. Modern displays are 4K (8 MP). Upscaling old photos to 4K resolution makes them look like they were shot today.

Low-res screenshots. A small UI screenshot blown up for a presentation goes from blurry to crisp. Our upscaler preserves text shapes much better than a stretch.

E-commerce hero images. A 800×600 product photo upscaled to 3200×2400 lets you show the full image at hero size without pixelation.

If you also want to clean up colour, sharpness, and exposure in the same pass, run the result through the Photo Enhancer after upscaling.

Limitations and what to expect

The AI is good but not magic. It can't recover detail that was never captured — you can't read text from a Times Square sign in the background of a 200×150 thumbnail. It can't undo motion blur (use the Sharpener for that). It can produce strange "hallucinated" details on extremely low-resolution faces (use the Face Enhancer for face-specific upscaling). And it costs 10 credits per call because each upscale uses real GPU compute — that's why the daily caps are tight.