What This Photo Restoration Tool Does
The AixKit Photo Restoration tool applies a layered enhancement pipeline to old, faded, or low-contrast images. It runs entirely on AixKit's server using PHP and the GD image library — no external AI model, no third-party embedding, no data sent off-site.
The four-step pipeline, applied in order to every uploaded image:
- Auto-levels (histogram stretch) — finds the darkest and brightest tones in each colour channel and stretches them to the full 0–255 range. This is the most impactful step for faded scans where everything has collapsed toward a grey mid-range.
- Gamma correction — applies a brightness curve that lifts midtones without blowing out highlights. Particularly effective on underexposed or dim photos taken in low light.
- Contrast boost — applies a mild contrast increase using a standard scaling formula to restore punch without crushing shadow detail.
- Sharpening — applies a Laplacian sharpen convolution kernel to improve edge clarity and fine detail. Kept moderate to avoid halo artefacts.
The result is saved as a PNG and returned alongside your original for direct comparison.
What Kinds of Photos Improve Most
This pipeline is specifically effective for a predictable set of image problems:
Works very well on
- Faded prints and scans — old prints lose contrast over decades. Flatbed scans often add a grey cast. Auto-levels corrects both in one step.
- Dim indoor photos — underexposed photos taken without flash or in artificial light benefit significantly from gamma lift and contrast.
- Scanned documents and certificates — boosting contrast and sharpening makes text and fine detail much more legible.
- Photos with washed-out colours — per-channel histogram stretch restores colour separation that fading has compressed.
- Low-contrast black and white prints — auto-levels gives particularly strong results on monochrome images where the tonal range has narrowed.
Works moderately on
- Slightly blurry photos — sharpening improves edge clarity but cannot reconstruct detail that was never captured.
- Slightly overexposed shots — gamma and contrast can help with mild overexposure, but severely blown highlights cannot be recovered.
Limited benefit for
- Heavily damaged or torn photos — the pipeline has no inpainting or fill capability. Physical tears, ink stains, and missing areas are outside its scope.
- Very low-resolution images — the pipeline does not upscale. A 200×150 pixel image will be returned at the same resolution.
- Heavily overexposed images — when highlight detail is gone, it cannot be recovered by brightness or contrast adjustments.
How Local Enhancement Differs from Neural AI Restoration
There is an important distinction between the enhancement approach used here and what neural AI restoration models do.
What this tool does (local GD pipeline)
Each pixel is adjusted mathematically based on the pixel values in the image. Auto-levels, gamma, contrast, and sharpening all work by reading and transforming existing colour information. No new information is invented — the pipeline can only work with what is already in the image.
What neural AI restoration does (not included here)
Models like GFPGAN, CodeFormer, and Real-ESRGAN are trained on millions of face and photo examples. They can hallucinate plausible detail — sharpening faces, filling in scratches, reconstructing missing texture — because the model has learned what these things should look like. This produces dramatically better results on damaged portraits but requires substantial compute, GPU infrastructure, and often a paid API.
Which to use when
- Use this tool for faded, dim, or low-contrast photos where the original information is present but suppressed.
- Use a neural AI model for physically damaged, scratched, or very low-resolution portrait photos where you need reconstructed detail.
Best Workflow for Restoring Old Photos
- Start with the best available scan. If restoring a physical print, scan it at 600 DPI or higher. Higher resolution gives the pipeline more pixel data to work with and produces a sharper result.
- Straighten and crop first if needed. Use the Image Editor to rotate and crop before restoring — the enhancement pipeline works on the whole image, so trimming off scanner borders first avoids them affecting the auto-levels calculation.
- Run the restoration. Upload here, wait for processing, and compare before and after.
- Download as PNG. PNG is lossless — your enhanced image is saved without any compression artefacts. If you need to share or publish it, you can convert to JPG at a later stage.
- Archive the result alongside the original. Keep both versions. The original is always valuable as a reference.
Common Mistakes to Avoid
- Uploading a JPG that has already been saved multiple times. Repeated JPG saves degrade the image before restoration even begins. If you have access to the original scan or RAW file, use that instead.
- Expecting colour reconstruction. The pipeline adjusts and stretches existing colours — it does not restore colours that have faded to a single-channel grey. Photographs that have fully desaturated to black and white will not become coloured.
- Processing already-sharp, well-exposed photos. The enhancement is designed for degraded images. Applying it to a well-exposed modern photo may over-contrast or over-sharpen it. The tool is best reserved for photos that need it.
- Ignoring the before/after comparison. Always compare both panels before downloading. If the enhancement has not improved the image noticeably, it may be that the photo is already at the limit of what a mathematical pipeline can recover.
Why PNG is Better Than JPG for Restored Images
The tool downloads results as PNG for a specific reason: PNG is a lossless format.
When you restore an old photo, the enhancement pipeline is working at the pixel level to recover tonal range, lift midtones, and sharpen edges. If the result is then saved as a JPG, the JPG compression algorithm introduces its own artefacts — blocking, ringing around edges, colour smearing — which can partially undo the enhancement, especially in the finer detail that sharpening was meant to recover.
Saving as PNG avoids this. The output file is larger, but the quality is preserved exactly as processed. If you later need a smaller file for web publishing, you can compress the PNG to JPG using the Image Editor with quality control, or a dedicated compressor, once the restoration work is complete.
Privacy and Local Processing
When you upload an image to this tool, it is processed on AixKit's own server. It is not forwarded to any third-party embedding platform, external AI API, huggingface Space, or cloud model host.
The previous version of this page embedded a cross-origin widget from a third-party host. That version has been completely replaced with a local PHP processing backend that runs entirely within AixKit's infrastructure. Your image data does not leave AixKit's server during processing.
Output files are stored temporarily in the server's outputs folder and are accessible only via a unique filename. They are not indexed, catalogued, or associated with your identity in any way.
Related Tools
- Image Editor — Rotate, flip, adjust brightness and contrast, apply filters, and export. Useful for pre-processing before restoration or fine-tuning after.
- Background Remover — Remove the background from a restored photo before using it in a design or presentation.
Frequently Asked Questions
Is my photo sent to an AI company or external service?
No. The image is uploaded to AixKit's server and processed locally using PHP and GD. It is not sent to any third-party API, neural network service, or external model.
What file formats are supported?
You can upload JPG, PNG, and WebP images up to 10 MB. The result is always downloaded as PNG for maximum quality.
Will the tool work on black and white photos?
Yes. Auto-levels, gamma, contrast, and sharpening all work on greyscale information. Black and white photos often show the most visible improvement, particularly scans with a grey cast.
Why does the enhancement look subtle on some photos?
If the original photo is already well-exposed and high-contrast, the pipeline has less room to make visible changes. The enhancement is most dramatic on images where the tonal range has genuinely collapsed — faded prints, dim scans, and underexposed shots.
Can the tool repair scratches or tears in old photos?
No. The pipeline adjusts colour and tonal values — it does not inpaint, fill, or reconstruct missing areas. Physical damage like tears, scratches, and ink marks require manual retouching or a neural inpainting model.
How long does processing take?
Most images process in under 10 seconds. Very large images (near 10 MB or high resolution) may take up to 20–30 seconds depending on server load. The progress bar will animate while the server is working.
Can I download the result as JPG instead of PNG?
The tool always outputs PNG to avoid introducing JPG compression artefacts into the restored image. Once downloaded, you can convert to JPG using the Image Editor with quality control if a smaller file size is needed.
What happens to my image after processing?
The output file is saved in the server's outputs folder with a unique randomly generated filename. It is not stored in a database, not associated with your account or identity, and is not accessible to other users. Original upload files are deleted immediately after processing.
