Oh man the last two months have been buusy. I have totally neglected this site + lost control of my lofting plans to chalk it full of research imagery from the past year. Nevertheless these plans still exist, not to mention a whole pile of documentation materials from the final thesis + exhibition to share with you all [random visitor.. yes you!]. Hopefully this will take place in the very near future, as there are 16+ hours worth of train travel in my immediate future = pretty good chance I’ll get a few things posted.

Until then, I offer a tease of some spontaneous algorithm studies for further visual possibilities that have yet to reveal themselves.. ooooh… ahhhhh.

The true HEADer_remix is coming, the true HEADer_remix is coming! soon, to a browser near you… keep your eyes peeled and your desire to dig beneath the image surface at ease… for now I present a sneak peak of the offerings:

_UPDATE

Just a few more tweaks and it will be ready.. oh man is this tool fun! Hopefully launching it by this weekend…

As previously described in semantic abuse, the logic of how this system works should leave the user hesitant to form any direct connections between the content of the text and the output of the image. I present here another image set that calls this very logic into question.

_DOUBLE PASS TEST

If it was only the color algorithm that produced a blue image, than I would have little to find surprising, SEAWATER created a very similar visual. It was when observing the b/w algorithm [alg_02] that I became particularly impressed.

_PRECONCEIVED NOTIONS ARE QUESTIONED

This continues the line of questions regarding semantics that must be asked of the tool itself, text2image, and the visual output being produced. OCEANWATER is a tangible word. When I conjure the mental image this text, the range of pictorial references are all very similar, in contrast to a commonly entered word such as Love.

When asked during a discussion the other day, “what do the images mean to you, what is the value of this system” - my initial response is that it provides a form of translation or visual representation. The most recent example of this would be my use of the tool last Sunday, inputting the phrase happy fathers day!

When describing this process to my professor of theory, an interesting concept began to unfold. If you imagine the phrase happy fathers day! [in pictorial form] what are those elements? Balloons, greeting cards, grown male figures – a predominantly blue color palette? Let’s see what Google Images has to say…

While these results are predominantly text based [providing an example of something difficult to communicate in pure visual language] - what if we repeat the process with that popular text2image entry, Love. Once again, bring the visual into your mind… this time I’ll refrain from offering any suggestions. As the short list of elements are formed, feel free to compare your imagination with that of a few websites with open search boxes :

- Google Image Search
- bing
- gettyimages
- flickr

If I haven’t already lost you to a sea of visual tangents, did the query of images reenforce your mental image? Let’s now take a look at what text2image will produce:

If this was the image that sat in your mind, please send me an email! If I were to try and draw a line of similarities between the variety of pictures found above and this latest one, it would likely remain on the topic of color. Which in many cases [as demonstrated with OCEANWATER] may be enough to deduct that the resulting picture is a true depiction of the text that was fed into the tool.

What about the case in which a picture fails to match a preexisting mental image [perhaps like the above example]? This may be where text2image has the ability to enable one to question their own preconceived notion of semantics. For the sake of this being a blog, [which promotes unfiltered thoughts to find themselves published on the internet], let us now make comparisons to that of an abstract painting [insert proper reference here... Picasso?] which is often displayed with an accompanying title written beside the work. In an abstract painting you are asked to rely on the fact that this title was the artists intent or inspiration while constructing the artwork. This is, of course, challenged with the introduction of an Untitled #____ naming scheme, further obscuring our text to image relationship.

One of the most interesting aspects of this whole discussion is observing the role text plays in the retrieval of these pictures. In the case of an image search engine, it relies on metadata [keywords which have been digitally attached to the image file] or by querying the filename itself. In both of these techniques the text sits on the periphery of the actual pictorial content. The pictured was observed and someone made a conscious and subjective decision to tag that information to the file. In the case of text2image, the text is the image and the image is the text! This is where the unique qualities of the digital medium come into play. By replacing the entire pictorial contents with the textual input provided, the text no longer sits on the periphery, but has become a core element in the visual outputs depiction.

While this relationship is explored through the immediacy of an interactive format, what information will the picture reveal if isolated and presented on its own?

*craaack… twist twist twist…* Hm that appears to be the sound of a whole new can [discussion] opening up. Time to walk away while enjoying and exploiting the riches of an informal writing medium.

_LET’S REMIX IT!

In the process of learning how to isolate the individual elements of an image using PHP I developed an interesting way to ‘remix’ the contents of the image body. IMAGE_REMIX finds obvious parallels with the CORRUPT™ work of Benjamin Gaulon, however on the website I have gone into detail describing exactly what brings this tool unto its own. Please visit I M A G E _ R E M I X for more information.

I decided it was time to start playing with some possible secondary uses for text2image outputs. While I probably find too many of the images beautiful and worthy of standing on their own, it’s my hope that they will also serve well as a raw material. I shall begin by utilizing a freeware application called SoundOfAnImage. As expected, most of the built in synthesized sounds are pretty cheesy, however it does have the amazing ability to export a midi file for use in any other program. This is something I shall begin to explore but to kick things off, I present, #20090218_005308599, #20090209_023015, #20090214_091343123, #20090422_190230977. I suppose it’s time to start digging into Steve Reich’s work…

+ Please listen with headphones to the full spectrum of frequencies produced…

_TREACHEROUS WATERS

I found this image last night while refreshing the website in search of the latest 15. Upon loading, a giant blue swatch stared me down, my eyes shifted to the caption, and the mysterious moment was broken with laughter. What funny things people decide to put into this system. Immediately my mind with its pattern establishing circuits went ‘WOW… tell me that is not a picture of an ocean?!’ - and it most certainly is. I imagine it as a sectional view, from the light colored surface on down to the dark depths of unchartered’ness. 

This is a dangerous game to play. From my understanding of how the whole system works + why it works, there is no room for semantic translation in the images; it has to be shrugged off as a coincidence. Nevertheless, there is a long slew of examples which could argue that a link between textual subject and image exists, backed up by a larger selection of text2image users who express similar notions in their observations. This is where I could/would/should hand the key off to a social scientist and ask for their professional opinion on the matter. In the meantime, I will have fun engaging in what I shall dub a textual fetish [article on that to come soon], and enjoy the fascinating images this tool produces. 

_REINFORCING THE POINT

Have a look at the b+w translation of alg_02 for the very same text:

_GOING TOO FAR

Finally, when the two are combined:

_TEXT2IMAGE VERSION 2.0!

So I finally got over my selfish ways and decided to share the latest text2image algorithm with the public. Finishing the new design, optimizing the code, and implementing functionality which will help me understand the usage of this tool, has kept me pretty busy up until this point. It’s now, most definiately, time to take my text beyond an outline phase, start writing my reflections in detail, while the thoughts [such as hoarding the new developments to myself] are still fresh in my mind.

_WHAT ABOUT VERSION 2.1? 

Hey, 2.0 JUST came out! …Yeah yeah.. still some little things to do. I would like to add an instant rating system, so that people could flag images they found to be aesthetically pleasing, thus increasing the speed at which I can wade through the database, find the best [subjective to me] images, and figure out just why they turned out the way they did…

Last night I was engaging in some productive procrastination, when I came across a particular work by Scott Blake [aka barcodeart man] known as the Free Chuck Close Art project. A painting is broken into a grid, extracting each of the many unique marks that make up the rasterization of of a portrait. They are then used to ChuckClose’ify any image provided by the user.

At the bottom of the page there is an animation of Muybridge’s The Horse in Motion run through this Free Chuck Close process:

Upon seeing it, I was immediately reminded of some earlier research I had done on the JPEG image header [using the very same raw material], where I had discovered a similar mark [or so I consider], which I dubbed the Chuck Close Aesthetic.

_AND THEN….

… I was inspired to work on a new tool, which could/should/shall repeat such an effect on any given image. I’ve drawing out a basic model of how it could work, and may begin work on it after bringing text2image to version 2.0. Until then, I’ll work on uploading some of the better examples of this research and share the results.

I have begun testing alternative algorithms for text2image. Here I’ve built a small tool that allows me to enter one textual input, while building 8 different images. From this display, alg_01 represents the current model in use, while alg_02, alg_03, alg_08 appear to be the most promising for new possibilities.

 

_MENTAL NOTES

I’ve got some decisions to make re: whether or not to publish working interactive versions of every new aesthetic I develop. By wanting to keep the latest output to myself, I’ve discovered an acquisitive personality that was previously unknown. Between personal intent and technical details, many questions remain…

+ what is more important, the image or the tool?

+ should there be exclusive content available only for showing in [the non-existent] exhibitions?

+ how do you maintain simplicity, while offering the user 3x the variety?

+ *insert next tough question here*