Log inUpgrade

AI Image Upscaler

Enlarge photos 2× or 4× with Topaz and get the detail back, or upscale video with SeedVR2 or FLUX 3. Drop a file to see the exact price before you pay.

Runs in the tool. You sign in to run it, see the exact price before you start, and failed upscales are refunded.

A black and white photo of a grandmother and a young boy laughing on a porch step, upscaled 4× from a 400 pixel scan
Before4×

Family photo, 1960s · 400 × 400 → 1,600 × 1,600. Drag to compare.

Real results

Drag across any image. Each pair is the file we uploaded and exactly what came back. The sources are shrunk, compressed copies of larger images. Use a mode to open the tool with it set.

Video, up to 4K

Every frame, rebuilt

SeedVR2 restores each frame and keeps motion steady between them, and FLUX 3 redraws them sharper at 24 fps. MP4 or MOV up to 20 seconds, and your audio comes along.

This clip
640 × 280 → 2,560 × 1,120
Length
4 s, 96 frames
See the Video Upscaler
Before · 280pAfter · 1120p
320 px1280 px

Phone video, 320p to 4×

320 × 568 → 1,280 × 2,272 · 4 s

Skin, hair, and the reflections on the bottle, restored frame by frame from a 320-pixel phone video.

Upscale an image in three steps

  1. 1

    Drop a photo or video

    Any photo, including iPhone HEIC, or an MP4, MOV, or WebM clip. Longer clips are trimmed to their first 20 seconds. Paste works too. Your file stays in the browser until you upscale.

  2. 2

    Pick a mode and scale

    For photos, pick a mode for faces, old photos, text, or art, then 2× or 4×. For video, pick SeedVR2 (2× or 4×) or FLUX 3 (1.5×, 2×, or 3×). The finished size and the exact price show before you pay.

  3. 3

    Compare and download

    Drag across your result to compare it with the original, zoom in on details, then download. Every upscale is saved to Assets.

Pay per upscale

Priced by the size of what comes back. The exact price is on the button before you pay · failed upscales are refunded · credits roll over

Photos, Topaz
By the megapixels of the finished image, in steps of 24 MP.
Video, SeedVR2
By every output megapixel of every frame, so a clip’s length, frame rate and scale set the price.
Video, FLUX 3
By every second of each output megapixel, at 24 fps whatever the source. Creative costs more than precise.
How Made pricing works

Upscale from code or an AI agent

The same model at the same price through the Made API. Preflight any job for free with /v1/cost, or pass a job_… ID as the source to upscale something you just generated. Assistants can do it over MCP.

curl -X POST https://www.made.app/api/v1/generations \
  -H "Authorization: Bearer $MADE_API_KEY" \
  -H "content-type: application/json" \
  -d '{"model":"topaz_upscale_image","upscale_factor":4,
       "medias":[{"role":"image","value":"https://example.com/photo.jpg"}],
       "params":{"model":"Low Resolution V2"}}'

Questions about upscaling

What does an AI image upscaler do?

It makes an image 2× or 4× larger in width and height and rebuilds detail as it goes: edges, textures, hair, and small text, instead of just stretching the pixels. Made runs Topaz precision models on fal. You get a new JPG or PNG, and your original is never changed.

How much does it cost?

It’s priced per image, in credits, by the size of the result: the price steps up with every 24 megapixels of the finished image. The exact price shows on the Upscale button before you pay, failed upscales are refunded, and unused credits roll over.

Which mode should I pick?

Standard fits most photos. Low-res rescue is for small, blurry, or heavily compressed images like thumbnails, old scans, and screenshots of photos. High fidelity is for photos that are already sharp and need to be bigger, since it changes the least. Text & logos keeps labels, signs, packaging, and screenshots legible. Art & CGI is for illustrations, anime, 3D renders, and AI art.

Should I use 2× or 4×?

2× doubles the width and height, so the image has 4 times the pixels. 4× gives 16 times the pixels. Use 4× for small images and 2× for images that are already large. Results can be up to 100 megapixels and 16,384 pixels on a side, and the tool greys out any scale that won’t fit.

What does face recovery do?

It is a Topaz pass that restores eyes, skin, and hair in faces. It’s on by default and costs nothing extra. It can smooth very fine skin detail such as light freckles, so switch it off when you want faces to stay closer to the original.

Can it fix a blurry photo?

It sharpens soft, low-resolution, and compressed images well. It can’t recover detail the camera never captured, such as heavy motion blur or a missed focus. An upscaler predicts detail, so check faces, small text, and logos before you use the result.

Which files can I upload?

JPG, PNG, or WebP images up to 30 MB and at least 32 pixels on each side. You can also paste an image from your clipboard. Results download as JPG or PNG, whichever you pick.

How long does it take?

Most images finish in 15 to 30 seconds. Very large results take longer: a 69-megapixel result took about a minute and a half. You can leave the tool, and the result waits in Assets.

Can I upscale videos too?

Yes. Drop an MP4 or MOV in the tool, or open the Video Upscaler. Clips up to 20 seconds go up to 4K with your audio kept, on SeedVR2 (2× or 4×) or FLUX 3 (1.5×, 2×, or 3×).

Can I upscale from code or an AI agent?

Yes. Call the Made API with the model topaz_upscale_image, the same one the tool uses, or connect Made to your AI assistant over MCP. Pass a job ID as the source to upscale something you just generated.

Where do my upscales go?

Every result is saved to Assets in your Made account, so you can download it again later.