Overview
Use AI Photo Restorer for soft detail, visible noise or blocky JPEG artifacts. Choose a matching Restoration Type, compare the PNG result with your original, and keep the copy if it suits your needs. Online processing requires sign-in and an internet connection.
Have a photo with blur, grain or blocky edges? Choose the problem you want to work on, compare the result, and save a restored copy.
When to use
- Choose this photo restorer for soft or blurred detail, grain or noise, or blocky JPEG compression. Pick the matching Restoration Type, or start with Auto Detect if you’re unsure which issue is most prominent.
- For a different task, use a separate tool: the photo colorizer creates a colorized PNG copy from a black-and-white photo, while the image upscalers offer enlargement. This restorer’s choices address blur, noise and JPEG artifacts.
How to use
Open AI Photo Restorer and sign in to its online workspace.
Add one .jpg, .jpeg, .png or .webp image, up to 25 MB. Under Restoration Type, choose Deblur for soft or blurred detail, Denoise for visible grain or noise, Fix JPEG Artifacts for blocky compression, or Auto Detect if you’re unsure. Model Width offers 32 (Fast) and 64 (Quality); choose the label that fits your priority. Restoration Strength ranges from 0.1 to 1 in 0.1 steps and defaults to 1. Review the visible controls and restrictions before continuing.
Use the workspace action and inspect the result. Compare faces, lettering and fine texture with the original; a cleaner image doesn’t establish what an indistinct feature looked like. If detail you want to keep looks smoothed away, try a lower Restoration Strength and compare again. If a downloadable result is offered, use Download to save the PNG and keep the original separately.
Settings reference
| Setting | Options | Default |
|---|---|---|
| Restoration Type | auto: Auto Detect deblur: Deblur denoise: Denoise dejpeg: Fix JPEG Artifacts | auto |
| Model Width | 32: 32 (Fast) 64: 64 (Quality) | 32 |
| Restoration Strength | 0.1 – 1 | 1 |
Formats & limits
| Detail | Requirement |
|---|---|
| Input formats | .jpg, .jpeg, .png, .webp |
| Output formats | PNG |
Access & cost
| Detail | Requirement |
|---|---|
| Processing mode | Online processing requires sign-in and an internet connection. |
| Online processing | Sign-in required; Internet connection required |
Examples & use cases
A family archivist preparing a digital copy of an old snapshot.
A soft family snapshot
The .jpg has soft facial detail. The archivist wants a separate version to compare without treating newly sharpened features as a historical record.
.jpg to PNG
- Restoration Type:
- deblur
- Model Width:
- 64
- Choose Deblur because the visible problem is soft detail. Choose 64 (Quality) if that Model Width label fits the task.
- Compare faces and fine texture in the PNG with the original. Don’t assume a clearer-looking feature is an exact reconstruction.
Expected outcome
A PNG result to compare with the original before deciding whether it’s useful for the archive.
A family historian organizing scanned photographs.
Grain in a family-history scan
A .png scan has visible grain across the image. The historian wants a processed copy to review while keeping the untouched scan for comparison.
.png to PNG
- Restoration Type:
- denoise
- Model Width:
- 32
- Choose Denoise because visible grain or noise is the issue. Choose 32 (Fast) if that Model Width label fits your priority.
- Compare faces, lettering and fine texture in the PNG with the untouched scan before deciding whether to keep the result.
Expected outcome
A PNG copy to review beside the original scan for the family-history project.
A community newsletter editor considering an older image.
Blocky compression in a photo
A .webp image has blocky compression around its edges. The editor wants to compare a processed copy with the source before deciding whether to use it.
.webp to PNG
- Restoration Type:
- dejpeg
- Model Width:
- 64
- Choose Fix JPEG Artifacts because the visible problem is blocky compression. Choose 64 (Quality) if that Model Width label suits the task.
- Inspect the edges and nearby fine texture in the PNG against the source before deciding whether the result is appropriate for the newsletter.
Expected outcome
A PNG result to review against the compressed source before use.
More ways to use it
- A genealogist reviewing a scanned portrait.: The portrait looks blurred, and the genealogist wants to see whether a separately processed copy is useful for reference. A PNG copy to compare with the scan. Treat indistinct facial details as uncertain, not as an exact reconstruction. Restoration Type: deblur.
- A family member organizing digital photo scans.: One scan has noticeable noise, and they want a processed copy without replacing the untouched scan. A separate PNG result to inspect beside the original scan before deciding which version to keep. Restoration Type: denoise.
Troubleshooting
The image file is larger than 25 MB.
Try this: Choose a supported image file no larger than 25 MB.
What to expect: The selected image is within the configured file-size limit; the workspace must still accept it for processing.
You’re unsure which visible issue is most prominent.
Try this: Choose Auto Detect under Restoration Type.
What to expect: Restoration Type is set to Auto Detect instead of one of the specific task choices.
The result appears to smooth away detail you wanted to keep.
Try this: Try a lower Restoration Strength, such as 0.7, then compare the result with the original.
What to expect: The next run uses a lower Restoration Strength setting; inspect the result to decide whether it better suits your task.
Privacy
Online processing uploads files. Completed results may remain available for later download. See Privacy Policy for current file-handling and retention details. Check the processing mode before starting.
Frequently asked questions
Can I restore multiple photos in a batch?
No. The tool accepts one image at a time. Add one supported image for each restoration run.
Should I scan the original photo again before restoring it?
If you still have the print, scan the whole photo with even lighting and avoid glare. Save an untouched scan so you can compare it with the processed result.
Will restoration add color to a black-and-white photo?
No. This restorer’s choices address blur, noise and JPEG artifacts. For a colorized PNG copy, use the separate photo colorizer.
What problems can this photo restorer address?
Restoration Type offers Auto Detect, Deblur, Denoise and Fix JPEG Artifacts. Choose the option that matches the visible issue. These listed choices don’t describe scratch, tear or stain repair.
Which image formats can I upload, and what format is the result?
The declared input formats are .jpg, .jpeg, .png and .webp. The configured output is PNG. Those declarations don’t establish support for every codec or bit depth.
Do I need to sign in to restore a photo?
This tool runs online and requires sign-in and an internet connection. Review the processing charge shown in the workspace before starting.
How are uploaded photos and results handled?
Online processing uploads files. Completed results may remain available for later download. See Privacy Policy for current file-handling and retention details. Check the processing mode before starting.
Does restoration recover exactly what a blurred or indistinct face looked like?
No. The result isn’t an exact record of missing detail. Compare faces and other fine details with the original; don’t treat restored facial features or lettering as an exact reconstruction.
Will the tool change my original image?
The configured result is a PNG copy. Keep your original separately for comparison, and use that original (not an earlier restoration) as the source for later versions you want to compare.
Can AI restore a severely damaged or very low-resolution photo?
The listed Restoration Type choices address blur, noise and JPEG artifacts; they don’t establish recovery of severe physical damage or missing detail. Compare any result with the source before relying on it.
Can restored photos be downloaded in high resolution?
The configured output format is PNG, but output dimensions aren’t specified here. Check the downloaded file’s dimensions before relying on it for a particular use.
What should I check before keeping a restored result?
Compare the PNG with the original, especially faces, lettering and fine texture. A cleaner appearance doesn’t prove that indistinct details are historically accurate.