RAWerse Alpha - one more tool to make your photos better


How does the magic work?

Learn More

About the RAWerse Alpha


How does the magic work? When you upload your image to the application, we run a number of smart and powerful algorithms right on your computer.

We appreciate your privacy - your data will never leave your computer.

The application is designed to work autonomously. It doesn't require any kind of internet connection.

As a bonus, you will never get denial of service or mainainance period of our servers.

Thus, as far as you don't use external computational power there is no sense in subscriptions or in regular payments of any kind. Just buy the app and it will be yours forever.

Simple and intuitive interface

Open application and drop some images you want to recover.
Wait a bit for images to be processed.
Look at the predicted quality histogram and additional information of processed image.
Drag image out of the application or click "Export PNGs ..." to save all at once.

System requirements

Minimum Recommended
CPU Intel® Core™ i3 3250 3.5 GHz or Intel Pentium G4560 3.5 GHz / AMD FX-4350 4.2 GHz Processor Intel® Core™ i5 2400 3.4 GHz or i5 7400 3.5 GHz / AMD Ryzen R5 1600X 3.6 GHz
RAM* 4GB 8GB
GPU** CUDA-compatible CUDA-compatible
Operating system 64-bit Windows 7, 64-bit Windows 8 (8.1) or 64-bit Windows 10 64-bit Windows 7, 64-bit Windows 8 (8.1) or 64-bit Windows 10
* depends on your photos
** only for full version with GPU support

Get Free DEMO Buy RAWerse Alpha

Features


Privacy

Your data will never leave your device

Machine Learning

Trained on millions of images we know how to fix yours

True RAW

You'll get *.dng file

Forever your

No subscriptions. No hidden payments

Let's delve into details for a bit. Imagine you have photos from your old camera, not even DLSR, just digital camera. Even comparing to recent shots from your phone they look not very impressive - mushy, have compression artifacts that look like squares along the image. You may have haze and overexposed regions as lamp bulbs or the sky. Also the images may be very small so they can't even cover your screen at full resolution.

Photo from Unsplash

When you try to enlarge your images in Photoshop or any other graphical editor, you realize that it's only getting worse. Whatever algorighm you are going to use the squares become monstrous, double edges appear near contrast objects and the overall image starts looking blurry.

Photo by Mario Calvo on Unsplash

There are couple of reasons of these problems. The first - when you convert RAW to JPEG you inevitable lose High Dynamic Range of the original image. It means that you're no more able to see what's behind overexposed regions because the data is already wiped. Well, it maybe even not your choice to convert RAWs but camera software. The second reason - JPEG is lossy compression, it removes your data trying to pretend there are nothing but flat and empty areas. That's why JPEG files are so small.

Photo by VietNam Beautiful on Unsplash

There are some comparison numbers for the image above.

Compression algorithm Image size
RAW / uncompressed 23.3 MB
PNG / loseless 1.7 MB
JPEG @ 100 / lossy 1.2 MB
JPEG @ 90 / lossy 354 KB
JPEG @ 80 / lossy 234 KB

It's true, when your image is already converted to JPEG there is no way to recover original data. But using deep analysis and deep neural networks you can statistically substitude some low-level textures, details, remove well-known artifacts. It's not going to be "fake" image, it will be just "better-looking" image, without artificial inserts.


Our application pipeline

For proper removal of JPEG artifacts our algorithm firstly infer possible rate of compression and then using this information remove all kinds of compression artifacts (squares, double edges, etc) and restore colors / details on corrupted regions if it's possible. To make it possible we run trained pixel-to-pixel neural network using Tensorflow framework.

In fact, the result is already here and we can simply save it as (loseless) *.png file. But to propose better RAW experience we can perform ML-dehaze using one more neural network. After that the processing ends and the result can be exported to *.dng file.

Digital Negative (DNG) is a publicly available archival format for raw files which are generated by various digital cameras. This addresses the lack of an open standard for raw files created by individual camera models and ensures that photographers easily access their files. (Adobe description)


We're really happy to share our results with you. All the beautiful images below were compressed with different JPEG quality for testing.

After, processed photo
Before, unprocessed image

Photo by Rhys Kentish on Unsplash (cropped, zoomed 200%, 50 / 100 JPEG quality)

After, processed photo
Before, unprocessed image

Art by Nadya Bratyakina (cropped, zoomed 300%, unknown JPEG quality)

After, processed photo
Before, unprocessed image

Photo by Cristian Morales on Unsplash (downsampled, cropped, zoomed 200%, 50 / 100 JPEG quality)

Looks magical, isn't it?
You can freely check results by yourself using image sources: [examples.zip]

Buy a full version


Free demo Full version
Export to PNG Yes Yes
"Processed with RAWerse alpha" Watermark Yes No
Export to DNG No Yes
Adjustable filtering settings No Yes
ML-powered dehaze No Yes
GPU support No Yes
Export PNG with original color profile No Yes
Free of charge (till august 2020):
€20.00 €0.00
(prices without VAT)

Purchase RAWerse Alpha for Windows 64bit (*.exe)

Let's Get In Touch!


Have something to ask or to tell? Please, share your thoughts with us