Upscale any image up to 4× — keep every detail crisp
Free AI image upscaler powered by Real-ESRGAN. Sharpens low-resolution artwork without smearing edges or inventing fake detail. Built for DTF transfers, gang sheets, and any print job where pixels matter.
Drop an image to upscale
PNG or JPG, up to 2 megapixels. Bigger images get resized down before processing.
How the AI image upscaler works
Drop your image
PNG, JPG, transparent or solid background. Up to 2 megapixels — anything larger gets resized down before processing so the AI has room to scale it back up.
Pick 2× or 4×
4× is the maximum and gives the sharpest result — best for tiny logos, low-res photos, or any source under 1000px wide. 2× is faster and lighter, ideal when the source is already decent and just needs a polish.
Sign in and download
A free account takes 30 seconds. The upscaled PNG exports at 300 DPI with correct metadata baked in — drop it straight into any DTF RIP, mockup tool, or gang sheet builder.
Why you should upscale before halftoning, not after
This is the single biggest mistake people make when prepping low-resolution artwork for DTF: they halftone first, notice the result looks blurry on a 12" tee, and try to AI-upscale the halftoned output. The upscaler — trained on natural photos — has no idea what halftone dots are. It tries to interpret them as image content, smearing the dots into blobs and inventing texture that wasn't there. The carefully-tuned dot pattern you spent time dialing in gets destroyed.
The correct order is: upscale first, then halftone. The AI gets clean source pixels to work with, you get a high-resolution image, and then you halftone at your target print size. The dots stay crisp because the halftone is the last transformation applied, not the first.
A good rule of thumb: if the effective DPI at your target print width is under 165, your image needs an upscale before anything else. A 600×600 logo printed at 12" wide is only 50 DPI — that's a blur on the shirt no matter what filter you apply. Run it through this tool at 4× first to get to 200 DPI, then halftone it. The difference is night and day.
When to use the AI upscaler
Customer-supplied artwork that's too small
The single most common DTF use case. A customer sends a 300×300 px logo and wants it printed 10" wide. That's 30 DPI — an unprintable blur. Run it through 4× upscale to get to 1200×1200 (still only 120 DPI but vastly sharper), or start from a larger source if you have it. The upscaled version cleans up jagged edges, smooths color transitions, and recovers detail the original lost during prior compression.
Old scanned logos and stock graphics
Logos from the early 2000s often live at 72 DPI screen resolution because that's all anyone needed when the asset was made. Modern Real-ESRGAN does a better job recovering letter forms and curves than traditional bicubic resampling — text edges stay crisp instead of going pillowy.
Photos and detailed illustrations
The Real-ESRGAN model is photo-trained and shines on detailed illustrations — skin tones, fabric texture, fur, foliage. If you're printing photo-style artwork on DTF, upscale the source before halftoning so the dots have a sharp base to grid against.
Mobile-screenshot artwork
Screenshots from Instagram, Discord, or text messages are almost always under-resolution for print. The compression artifacts and color banding in those sources get cleaned up substantially by the AI upscale — it's rarely perfect, but it's much better than what you started with.
Upscaled? Now make it print-ready.
Upscaling fixes resolution, but DTF needs more — clean transparency, no white haze on gradient edges, no stray specks. Make Print Ready scans your design for every other DTF issue and fixes them in one click. Free, no credit card.
AI image upscaler FAQ
Is this AI image upscaler really free?▸
Yes. Upscaling on this page is free with a DTFWiz account — no credit card, no trial expiry, no watermark on the output. Free accounts get 15 AI operations per day shared across upscale, background removal, and vectorize. Pro and Business accounts get higher daily limits.
What model does this use?▸
Real-ESRGAN — specifically the nightmareai/real-esrgan checkpoint running on Replicate. It's one of the most widely-used open AI upscalers, photo-trained, and known for clean edges on logos and text.
Does this work on transparent PNGs?▸
Yes. Transparent edges stay transparent. The model is run on the visible pixels and the alpha channel is preserved through the round-trip. The output is a transparent PNG you can drop straight onto a colored shirt mockup or gang sheet.
Why does it cap at 4×?▸
Real-ESRGAN is trained at 2× and 4× scales. Anything higher would have to chain multiple passes, which amplifies artifacts and rarely improves the result. If you need more than 4× from a small source, run the 4× output back through the tool a second time — but the second pass usually adds less than people expect.
What's the maximum input size?▸
Inputs over 2 megapixels (roughly 1500×1300px or so) get resized down before processing. This is a memory limit on the GPU side — the model runs out of VRAM on huge inputs. If your source is already that large, you probably don't need to upscale it; resize it down for the print width you want instead.
Can I use the upscaled image for DTF transfers?▸
Yes — that's exactly what this tool is built for. The PNG exports at 300 DPI with correct metadata for any DTF RIP. After upscaling, send the result straight to our halftone tool, Make Print Ready scan, or the editor for further cleanup.
Should I upscale before or after halftoning?▸
Always upscale first. AI upscalers smear halftone dots — they were trained on photos and try to interpret the dots as image content. Upscale your clean source first, then halftone the high-resolution result. Never the other way around.
How long does an upscale take?▸
Usually 10–30 seconds depending on source size and current load on the GPU. Very small sources go faster, very large ones take longer. A loading indicator runs during processing.
Will this work on vector files like SVG?▸
No — and you wouldn't need it to. Vector files have infinite resolution by definition; you can export them at any size without quality loss. This tool is for raster files (PNG, JPG) only. If your customer sent a vector logo, just export it at the target print size from your design app.