Blog

A Remini alternative that works in your browser

If you are searching for a Remini alternative, you probably want sharper faces and cleaner old photos without installing anything or signing up for a monthly plan. This guide explains how Airpic restores old photos online, how to keep the face looking like the real person, and what it costs you in time and privacy. It also tells you where the tool will not help.

What people actually want from an alternative

Nobody types a search like this out of curiosity. It usually means something got in the way. The details vary, but the list of needs is remarkably consistent, so it is worth writing down before looking at any tool:

  • No subscription. Many people restore a handful of photos a year: a grandparent's portrait for a funeral, a wedding photo for an anniversary, a box of prints after a house move. A monthly plan for three photos feels wrong.
  • No app to install. You want to open a page, drop the photo in and get a file back, on a phone or a laptop, without granting an app access to your whole camera roll.
  • A face that is still the same person. This is the big one. An AI enhancer that produces a sharp, attractive face that is no longer your grandmother has failed, however good it looks.
  • Seeing the result before paying. You want to judge the output on your own photo, at full size, before any money changes hands.
  • Control. A slider, a choice of steps, a way to undo. Not a single button with a single answer.
  • Clarity about your files. Family photos are personal. You want to know where they are processed, how long they are kept and whether they end up training a model.

Whatever service you are coming from, check its own terms for how it handles each of these points: they change, and they are not the subject of this article. What follows is how Airpic answers each need, with the settings you will actually touch. If you only want the short version, the tool is restore old photos online, and it runs in any modern browser on a phone or a computer.

Keeping the likeness: the fidelity slider and reference photos

Face restoration models work by recognising a face and rebuilding the missing detail: eyelashes, skin texture, the edge of the lips. When the original is badly degraded, the model has very little to go on, and it fills the gap with what a typical face looks like. That is how a restored portrait ends up looking like a stranger.

Airpic's old photo restorer gives you two ways to fight this.

1. The face fidelity slider. When the tool rebuilds faces, the Strength slider controls how much of the restored face is blended back into your photo. Before blending, the restored face is aligned to the original using its key points, and its colour is matched to the skin tones of your photo, so lowering the slider does not create double outlines around the eyes and teeth. The default is 60%. As a rule of thumb from our own tests:

  • For a reasonably sharp print with small faces, 80-100% gives the cleanest result.
  • For a very degraded source, such as a photo of a photo or a compressed screenshot, 40-60% keeps the person recognisable. Above that, the model starts inventing a generic face.

2. Reference photos of the same person. Under the options there is a block for reference photos. Add one clear photo of the same person (one is enough; you can add up to 3). With references, the job works differently: the tool generates several candidates guided by the identity in your references, aligns each one to the pose of the original, and shows them in a strip with a likeness score. You pick the one you recognise, and the comparison view and the download switch to it. The slider becomes a Fidelity control, set to 85% by default, which blends the guided face with the original features so the expression, a smile for example, survives.

Choose references with care. A front-facing photo in good light, with no hands near the face and no sunglasses, works best. The tool deliberately crops references tightly so it does not copy hands, hats or clothing from them.

An honest warning. Running the same photo through face restoration two or three times does not make it look more like the person. Each pass adds more of the model's idea of a face, not more of your relative. If the first result is off, lower the slider or add a reference instead of repeating. And when a face in the original is only a few dozen pixels wide, no model can recover features that were never captured: the tool shows a note when faces are very small, and the result should be read as a plausible reconstruction, not a document. Our guide on what can really be recovered from a pixelated WhatsApp photo goes into the limits in more detail.

The whole photo, not only the face

A good restoration is not a sharp face pasted onto a blurry, scratched background. Airpic treats the photo in stages, and you can see each one.

  • Straighten. Tick the straighten option and the tool detects the lines in the scan and rotates it. In our tests on 19 September 2026, a scan tilted by 2.5 degrees was straightened automatically with a correction of 2.48 degrees. If the photo has no clear straight lines, the tool leaves it as it is and tells you so.
  • Repair damage. Tick Repair and a mask editor appears: paint over the scratches, creases, tears or ink marks. The painted areas are filled from their surroundings before anything else happens. There is no automatic scratch detection, so this step is manual, but painting a crease takes seconds.
  • Clean. Grain, dust and the softness of an old print are cleaned up and the photo is enlarged, at ×1 or ×2.
  • Faces. The step described above, with its slider.
  • Colour. Optional: a black-and-white photo can be colourised in the same job.

When the job finishes, a row of chips lets you click through each stage and see what it changed in the comparison view. In our tests on 19 September 2026, an old photo of 1000 × 1289 px was restored in 9 seconds across three stages: clean-up, faces and colour. That is with the service already warm; the first job after a quiet period takes longer while a GPU starts.

Sometimes the restorer is not the right tool, and a more specific one does a better job:

The input can be JPG, PNG (including 16-bit and transparency), WebP, TIFF, HEIC or AVIF, so a scan from a flatbed or a photo straight from an iPhone both work. The orientation stored in the file is applied and the colour profile is kept. The restorer accepts photos up to 20 megapixels; scans larger than that should be reduced first.

AI photo enhancer without a subscription: cost and privacy

Airpic has no account and no subscription. You never give an email address or a password. This is how the money side works:

  • Processing is free. You upload, process and look at the result in the comparison view without paying anything. You can try every setting on your own photo first.
  • One full download a month is free. For many people restoring a single portrait, that is the whole bill.
  • After that, credits. You buy a small pack only if you need more downloads, and credits never expire. They are kept in your browser as a code, with no account behind them. Current prices are listed on the photo restoration page, and our article on trying a photo tool before you pay explains the logic.

Privacy is part of the same design:

  • Your photo and your reference photos are processed on a GPU in the European Union.
  • The original is deleted when the job finishes. The result is deleted at most one hour after you download it, or one day if you never do.
  • Your images are never used to train models.

Compare this with sending a family photo to a general AI chat assistant. A chatbot generates a new image that resembles yours, usually at 1-2 megapixels, and it can change faces, text and proportions along the way; the conversation may also stay in your history. A restoration model keeps the original pixels and adds coherent detail. Our article on why enhancing a photo with an AI chatbot changes your picture shows the difference, and AI image processing in the EU covers the data side.

Restore old photos online without an app: a two-minute test

The quickest way to judge any tool is to run your hardest photo through it. Pick one old print that matters to you, or a compressed photo someone sent you on WhatsApp. Then follow these steps.

  1. Get the photo in. On a computer, drag the file onto the page or paste it with Ctrl+V (Cmd+V on a Mac) after copying it. On a phone, tap to choose a file, or use the Take photo button that appears on the restore tool on touch screens. If you photograph a print, do it in daylight, flat, without flash, filling the frame.
  2. Choose the size. ×2 for a print or a screen; ×1 if the scan is already large.
  3. Set the faces. Leave face restoration on. For a clean print keep the slider at 60% or raise it; for a screenshot or a photo of a photo, start at 40-50%.
  4. Optional steps. Tick straighten for a crooked scan, tick Repair and paint over scratches if there are any, and tick colourise if you want colour.
  5. Add a reference if the face matters most. One clear photo of the same person is enough. Expect the job to take noticeably longer, because several candidates are generated.
  6. Process and compare. Drag the divider in the before/after view, or on a phone use the button that switches between before and after. Click View at 100% to inspect eyes and skin at real size, not a reduced preview. Use the stage chips to see what each step did.
  7. Adjust or download. If the face is too smooth or not quite right, use Redo with other settings, lower the slider and run it again; processing again costs nothing. When it looks like the person, download the full file. On a phone you can share it straight to a messaging app.

If your source is a scanned print, our guides on how to restore old photos with AI and restore a scanned photo cover scanning resolution and file formats. For a portrait that exists in only one copy, read restoring an old family photo when only one copy is left before you start.

Ready to try it on your own photo? Open the Old Photo Restorer, drop in the picture and look at the result before deciding anything.

FAQ

is there a free app to restore old photos without subscription

Airpic is not an app to install: it runs in the browser on any phone or computer. Processing and previewing are free, you get one free full download a month, and after that you can buy credits that never expire. There is no subscription and no account.

how do i stop ai from changing a face when restoring a photo

Lower the face fidelity slider to 40-60% when the original is very degraded, and add one clear reference photo of the same person if you have it. Repeating face restoration several times does not help: each pass adds more of the model's idea of a face.

can i restore an old photo on my phone without downloading an app

Yes. Open the Old Photo Restorer in your phone's browser, choose a photo or take one with the camera button, process it, compare before and after, and download or share the result. Nothing needs to be installed.

are my family photos kept after restoring them online

On Airpic, no. Photos are processed on a GPU in the European Union, the original is deleted when the job ends, and the result is deleted at most one hour after you download it, or one day if you do not. Images are never used to train models.

We measured it

  • A 1000 × 1289 px old photo restored in 9 seconds, in three stages: clean-up, faces and colour. (19-09-2026)
  • A scan tilted by 2.5 degrees was straightened automatically (2.48 degree correction). (19-09-2026)

Measured on airpic.app with the GPU already running. If nobody has used it for a while, the first job takes 40 to 90 seconds longer to start.

Restore an old photo now

Keep reading

Which AI models does Airpic use, and under what licence?

Which AI models does Airpic use? Per tool: Real-ESRGAN, GFPGAN v1.4, BiRefNet, DDColor, LaMa and Whisper, with licences, size limits and where they run.

Image upscaling API with no sign-up

Airpic's image upscaling API needs no sign-up, API key or card: a credit code is your bearer token. What you can call, how you pay per download, the limits.