FeedBlast – Rhizome Proposal

Introduction

FeedBlast is an experiment in directed generative video. It takes the form of a web-based application that creates an audiovisual mashup of web content using items from a collection of RSS feeds, also known as an OPML file. The goal of the mashup is to present a densely-packed summary of a users unread feed items in the smallest amount of time possible. The resulting Quicktime movie can be downloaded, viewed, and shared. Watching a FeedBlast is not about comprehension, but about gaining a general understanding of the information landscape that informs and directs later review of the feed information.

The FeedBlast Process

1. The user goes to the FeedBlast site and provides an OPML file.
2. The user sets several preferences about how the video should be generated, including how long each clip should be.
3. The application parses the OPML file and all the feeds within and gathers all of the items that were added since the last time a FeedBlast was generated.
4. The text of each item is analyzed with Topicalizer – a service that provides a variety of useful information on a given text including keywords and a short abstract.
5. The abstract is converted to speech using the AT&T text-to-speech software. A voice font is chosen from a list of 12 based on which voices were used before.
6. Any media items that appear in the items are gathered. This includes images,videos, and audio. Videos are converted to a usable format with the aid of the Vixy.net transcoding utility.
7. If there is no media in the body of the item, the application will follow any links within the text and try to gather media linked on the sub-pages.
8. If no media is found in the first level of links, the application will use the Yahoo Image, Audio, and Video search APIs to find appropriate media, using words and phrases deemed to be important according to the Topicalizer analysis as query terms.
9. The pieces are mixed together using the QuickTime API and the editing techniques shown below.
10. The video is made available for download.

The Cocktail Party Effect
Advances in filtering and delivery of syndicated web content have significantly increased not only the quantity of information a person can access, but the relevance of the information, and this trend will only continue. Technologies that aid in digesting information have not kept pace with the related delivery methods, leaving many people completely overwhelmed. The goal of FeedBlast is to remedy this problem using a phenomena colloquially called The Cocktail Party Effect, which is the ability to focus one’s listening attention on a particular audio source among a cacophony of background noise. By barraging the user with audio-visual information rather than just a long page of static text and images, I hope to utilize the Cocktail Party Effect to allow a user to focus their attention only on what is most interesting to them, and therefore allow greater information intake.

An Example

The following video was produced using the top 20 new items in my RSS reader on March 26, 2007, at 5:26 PM. FeedBlast will create something very similar.

Production Budget

** I am requesting $3,000 from Rhizome to help with the intern salary. I have applied for a NYSCA grant to cover the remaining portion.

Rough Production Timeline

FeedBlast will build upon the work of Jason Freeman’s open-source iTunes Signature Maker. This will significantly shorten the development process. The project is divided into three parts: the user interface, the media hoarder, and the mixer. Development will start on July 1 and continue until November 1. I will be receiving design and editing assistance from Jerry Juarez – a Senior Production fellow at Eyebeam.

About Jeff

Jeff makes things out of the Internet. His projects include both art pieces and software development, but most fall somewhere in between. Most of his work involves generative systems that use pre-existing content. He’s not an artist or developer as much as a hacker. He received his MS from the Digital Media program at Georgia Tech and in 2006 and then joined Eyebeam Atelier as a production fellow.

http://www.jeffcrouse.info/cv.html

Work Samples

1. Interactive Frank is a program that uses resources from the web to construct a narrative based on an initial sentence entered by the user. It was written in Java.
2. Switchboard is library for Processing that provides a conceptual level interface to a variety of web and network services.
3. Earthify takes a page of Craigslist posts and maps them onto Google Earth.
4. Real-Time Art is my masters thesis at Georgia Tech. It explores common strategies and qualities of art that uses live data sources.