The ''Deep-Picture Converter, Version 1.0'' (only for Windows) available on the Download page of this website

This is how it is still looking today (6th of April 2022). Some little improvements in style and matters of user friendlyness have caused a pretty nice software that nearly everyone should easily be able to use.

The design completely changed

Nilesh, the programer of the software, created a very nice update. So now it looks pretty cool, how I would say.Sadly the colours got a little ''facebook-like''.

One could say the software of the Deep-Picture Converter already has something like a history. Since the completition of the first prototype on the 17th of October, 2021 there have been a lot of updates. Problems had been solved, improvements in colour have been made, technical innovations took place and last but not least the converter got a modern design. So on the 23rd of January 2022 the first ''finished'' Software was released (still only for Windows) and could be loaded in the download-section of this website.

HERE you can find the software for download.

Try to imitate this effect with an existing picture editing tool - you'll fail, this is all new!

Update of the Prototype

The colouring has been improved, it is now exactly calculated. Also exponential growth is now possible. On the upper panorama you can see an example of exponential pixel-growth. On the one below the old, linear growth with its different character.

A panorama like this easily shows the difference between the two options that are already possible.

The colouring

Beneath this text you can find the comparism of the three possible colouring options. They can be calculated by examining the RGB-values, picked by determining the most occuring colour or randomly taken from the mother picture - which creates different results for each conversion.

Random occuring colour

Most frequent colour

Average colour (RGB-calculation)


PNG - 61,6 MB; JPG - 13,7 MB; HEIF - 11,9 MB

Save Memory

The aesthetical sense or nonsense of the converter is a question of personal view. But what is with the technical question? I have often talked about the possibility of saving lots of memory space, what could especially be useful when dealing with larger pictures or e.g. videos, but how are the facts?

I took a 10'000 x 5'000 pixels example picture and converted it into many different types of files. You can see Frankfurt (City in Germany), it's the view from the Main-river. The picture is lizense free and was taken from

Also mind ''LGBTQ+'' in the download section of this website - An 50'000 x 6'260 pixels example to show how much ''.heif'' comprimizes

Just these days I have discovered the advantages of the HEIF-file type for usage in combination with the deep-picture converter. When dealing with blank photographs, you won't see much difference in efficiency compared to jpg. But when dealing with deep-pictures and their characteristic squares, the HEIF format beats all. It is a lossy compression, but all of the details are being kept, the boat and it's name are as clear as in the original and compared to the converted deep-picture PNG file the human eye will hardly find a difference.

On the upper picture you can see the original file, while the one below shows an exponential deep-picture of the photo.

Deep Picture

PNG - 5,36 MB; JPG - 2,86 MB; SVG - 8,44 MB; HEIF - 519 kb

Keep the general view

The effect doesn't have to dissolve everything inside a pixel-matrix. A general resolution, that still keeps the picture high resoluted on most devices, can be kept easily.

Keep the details

There are a lot of pictures, where there is actually just one tiny part that consumes all of the viewers attention. These parts are the only ones that users tend to zoom in. So why should the whole picture be highly resoluted? And even if your not interested in saving space, which may be not so important when dealing with small pictures, you can set an excellent focus. Secondary stuff is beeing thrown out.

No sharp edges

A Deep Picture isn't just splitted into high and low resoluted disconnected parts, the borders are smooth. The pixels grow steadily from the center to the pictures corners. One can choose between linear and exponential growth (This one is exponential). For both of these options, the pixel-growth can be adjusted very precisely. Is there a point, in which the resolution suddenly changes? - No, everything is smooth! In the example below you can see the net structure, that is because I took a snapshot of the SVG-file.

The linear net-structure

Here you can see the linear net-structure. Row by row the pixels grow by an absolute value, in this case it is ''1''. This algorithm is possible since the first prototype got finished on the 27th of October, 2021.

The exponential net-structure

Here you can see how the exponential net-structure is beeing built up. Every of those little black fields is creating a coloured ''pixel'' in the converted Deep-Picture. I like this one more than the linear structure, it is more beautiful and creates more useful results, how I think.

Extend the high resoluted area

Of course the center doesn't have to be a single pixel, it can be extended as much as the user want's. In this case it is important to see all of the motorbike in high resolution, so the pixel growth starts beyond that area. Here I have choosen the linear structure for conversion.

Create an artwork

As you may have already been reading, you can create fantastic artwork when using the transformation in a useful way. Beauty lies in the eye of the viewer, some people think this is cool - some don't. Especially fans of pixel art seem to like this very much, as I noticed when presenting it to the social networks.

Possible adjustments

(1) Selecting a picture

Of course users can choose the picture that they prefer to use for convertion, PNG and JPG are supported inputs. You could try other file-formats, they could possibly also already work.

(2) Choosing the kind of pixel-growth

The user can select, weather a linear or exponential growth shall be used for conversion.

a) Linear – The pixels are growing row by row by an absolute number.

b) Exponential – The pixels are growing row by row by a typed in factor, while the resulting pixel size of each row is determined by using roundingmath.

(3) Choosing the colouring

The user can select, how the colour of the deep picture shall be originated.

a) Most frequently – The most often occuring colour from the mother-picture within a specific area determines the colour of the whole square of the resulting net structure.

b) Random – One random colour from the mother-picture within a specific area determines the colour of the whole square of the resulting net structure.

c) Average – The colours of the resulting net structures are calculated through the RGB- values by evaluating all pixels from the mother-picture.

(4) Typing the size of the center

Users can type the size of the high resoluted center (Detail area), it has to be a natural number (e.g. 1;3;400), which will determine the height and the with of the center square measured in pixels.

(5) Typing the size of the first row

The users can type, what resolution the first row, that is surrounding the center, should have. It has to be a natural number that is equal or smaller than the size of the center. (e.g. if the center measures 10x10 pixels, the first row has to be ''10'' or smaller.) Usually this is 1 - like the center is, or 2 - which is creates a difference depending where the growth actually starts. But experiments can be cool...

(6) Typing a value for the growth

Select how fast the pixels should grow by typing a so called ''pixel-incrementor''. For the linear growth it should still be a natural number (e.g. 1 or 2), for the exponential growth, it has to be a factor which is bigger than ''1'' (e.g. 1.03 or 2.5). Users shouldn't use a comma, they should use a dot.

(7) Choosing the center or manually typing the values for X and Y

Users may select the origin of the center by clicking on the button. A window will open where you can set a red center mark. You then have to save this location. Alternatively users can set the location of the center by typing the specific value for x and y (number of pixels measured from the upper left corner). The values x=100 and y=100 are default.

(8) Choosing, if the result should open up after conversion

Users may click the box, if you want the result to be shown up after conversion. In Version 1.0 the input picture sadly also opens up, when this option is selected.

(9) Selecting to which destination the converted files will be saved

Users can choose where to save the converted SVG and PNG files, as well as the net-structure, after the conversion.

(10) Choosing the language

You can choose between English and German by clicking on the country symbols in the upper right corner of the GUI.