Stepok Picture Enlarger| Stepok Image Lab

Stepok Picture Enlarger 3.2

Enlarge photo faultless with the help of AI technology

The Stepok Picture Enlarger is a professional software that allows you to enlarge and resize images with ease. Powered by the deep learning technology(AI), it provides outstanding results comparing to traditional resize methods such "bi-cubic". The edges and details of the object are kept sharp and natural looking while introducing an innovation way to add details in high resolution space include fur, fabric, hair and so on. Stepok Picture Enlarger is a perfect image resizer for photo, web image, vector style graphics, scanned text and hand-draw image.

Our quality software operates on Windows XP/7/8/10, supports single or batch processing, all common image formats and RAW files. Batch processing and Photoshop plugin included.

Runs on Windows XP/Vista/7/8/10, Max size to 20,000x20,000 on 64bits system, 10,000x10,000 on 32bits system
at least 4GB RAM, SSE2 support

Latest updates, v3.2 Algorithm improving.

How it works?

Like other AI algorithms, numbers of pictures are collected and organised into pairs of the original picture and a low resolution copy. Then a neural network is created, followed by a training that teaches the neural network to use the original picture as reference, generating realistic results from the low resolution copy.

Tiny Neural Network - AI runs on every computer

There are some AI based super resolution (image enlarger) products on the market and they provide outstanding results. To achieve the best quality, they are using a huge scale of neural network. Therefore, a powerful GPU is required to run the programs.

However a majority of home and office computers only have their integrated GPU. The speed of a integrated GPU is much slower than an expensive standalone GPU card (10-1000x). This makes it impossible for home and office computers to run those programs, because it is too slow.

In order for everyone to benefit from the latest technology, Stepok Image Lab spend years on developing a very tiny scale of network that runs on CPU instead of GPU. The algorithm we now have has a quality that's better than traditional algorithms. It's even better than some simple AI solutions, and close to top AI solutions. It is a tradeoff that exchanges some quality, with the ability of running on almost any computer.

Examples

Drag the area in middle-bottom of the picture to move the line.