Upscale JPG Photos — Enlarge 2x or 4x and Fix JPEG Artifacts
Upscale JPG and JPEG images with AI that simultaneously enlarges and removes compression artifacts. The AI reconstructs detail lost to JPEG compression, producing cleaner results than the original at a higher resolution.
Free local tracing needs no account. AI tools use credits and upload files for processing.
Explore features and examplesFrom your image to editable paths
- 01
Open your file
PNG, JPG, WebP, AVIF, GIF, or BMP up to 10 MB in the free workspace (the AI converter takes PNG, JPG, WebP, or AVIF). Existing SVGs can be edited in the free workspace.
- 02
Trace and inspect
Preview the shapes and colors. Clear logos and flat artwork usually trace more predictably than photos.
- 03
Make it yours
Edit the SVG, compare optional optimization, and download your result.
Working with your SVG
Recraft Crisp Upscale processing
Recraft Crisp Upscale processing
In Your Browser
Upload a PNG, JPEG, or WebP file and download the result when processing finishes. Nothing to install.
PNG, JPG, and WebP input
PNG, JPG, and WebP input
No Watermark
Download the result without a watermark, then check it before you use it. Results depend on the source image.
Preview before download
Preview before download
Multi-Tool Platform
After processing, use our other AI tools — upscaling, restoration, vectorization — all in one platform with shared credits.
AI conversion features
- Process JPG/JPEG files directly
- Recraft Crisp Upscale processing
- PNG, JPG, and WebP input
- Enlarged raster output
- Works with photos and graphics
- Preview before download
- No software installation required
- Works in any modern browser
- Commercial use allowed
- Pay-per-use — no subscription
- 1 free credit after you verify your email
Questions about your file
Free local tools. Optional AI credits.
Local tracing and SVG editing need no credits. AI conversion includes one credit after email verification; additional packs start at $9.99.
Why JPEG Upscaling Needs Artifact-Aware AI
JPEG compression works by dividing images into 8x8 pixel blocks and discarding high-frequency detail that the human eye is less sensitive to. This is fine at high quality settings, but at moderate to low quality, visible artifacts appear: block boundaries become visible, sharp edges develop "ringing" halos, and smooth gradients break into stair-stepped bands. When you naively upscale a JPEG, these artifacts get enlarged right along with the image — a blocky 500px image becomes a blocky 2000px image.
AI-powered upscaling solves this because the model has been trained on millions of image pairs — degraded JPEGs alongside their clean originals. It learns to distinguish real image detail from compression noise and reconstructs what the original likely looked like before compression destroyed it. The result is not just bigger, but genuinely cleaner than the input. Fine textures like hair, fabric weave, and skin pores are reconstructed rather than smeared.
This dual capability — simultaneous denoising and super-resolution — makes AI upscaling particularly valuable for JPEGs. You effectively get two operations in one: the image is cleaned up and enlarged at the same time. Running these separately (denoise first, then upscale) produces worse results because denoising alone tends to over-smooth, removing real detail along with artifacts.
Pro Tips for Better Results
Upload the highest-quality JPEG version you have
Every generation of JPEG compression destroys more detail. A photo downloaded from Instagram (re-compressed at perhaps quality 70) will upscale noticeably worse than the original from your camera (quality 92-95). Check your camera roll, cloud backup, or email attachments for the earliest version of the photo.
Consider saving the upscaled result as PNG, not JPEG
After the AI has painstakingly reconstructed detail and removed artifacts, saving back to JPEG re-introduces compression. If the image will be edited further or used in print, save as PNG to preserve everything the AI recovered. Only save as JPEG if you specifically need a smaller file for web delivery.
Use 2x for social media photos, 4x for print
Social media images are typically 1080-2048px on the long edge, so 2x gets you to 2160-4096px which is plenty for any screen. For print at 300 DPI, you need more: a 1200px JPEG only prints at 4 inches. Use 4x to get it to 4800px, which prints at 16 inches — enough for a large poster.
Check for color banding in sky and gradient areas
JPEG compression is hardest on smooth gradients, turning them into visible color steps (posterization). After upscaling, zoom into sky areas and skin tones to verify the AI smoothed these out. If banding persists, the source JPEG was likely compressed below quality 40 — at that point, even AI has limited material to work with.
How AI Removes JPEG Artifacts While Upscaling
JPEG artifacts follow predictable patterns tied to the 8x8 DCT (Discrete Cosine Transform) block structure. Block boundaries create grid-like lines, quantization of high-frequency coefficients causes ringing near sharp edges, and aggressive chroma subsampling (4:2:0) reduces color resolution to half the luminance resolution. Our AI model recognizes these specific degradation signatures and reverses them during upscaling. It reconstructs high-frequency detail that was discarded during compression, smooths block boundaries while preserving real edges, and restores full chroma resolution. The model was trained on JPEG images compressed at quality levels from 20 to 95, so it adapts its artifact removal intensity based on the severity it detects in your specific image.