Sunday, 30 October 2011

Extracting the stylistic elements from anything

I just posted about the identification of repeating elements from documents, particularly maps, to be able to apply the stylistic quality of that document to other documents. I mentioned some other types of document that this could be applicable to, but the more I think of it, the more it seems to me that this concept can be applied much more widely. As long as we have a way to structure the information we are getting.

I would be as bold as to say we can capture the stylistic quality of any human output, and be able to use that style for future output. And what is most interesting is that information technology gives us the power to extract, store and apply that stylistic data. And thereby give any untalented individual the ability to produce something in the style of an existing human output.

Let's take some examples:

Visual art - extract the style of how people, food, a room and a table look from Michelangelo's The Last Supper, and allow people to produce works of art in that style

The written word - extract the writing style of Shakespeare from across his plays (his vocabulary, sentence order, etc), and provide translation software from modern English into "how Shakespeare would have written it"

A voice - extract the accent of JFK's voice from his speeches, and apply to any written text with voice synthesis. A further possibility (and complication) would be to combine that with generated video of JFK. This method will allow deceased actors to continue to act after they are dead.

Another interesting aspect of being able to extract that stylistic information is the possibility of hybridising (multiple) styles, mutating styles, and applying iterative artificial selection to styles.

No comments: