TextFM: Open Broadcasting System. An interview with Graham Harwood and Matt Fuller

TextFM is a piece of software created by Graham Harwood and Matt Fuller. Using TextFM, anyone with access to a mobile phone can send a message to a specified number. A computer receives the message and reads it out using a text-to-speech program. The messages are then broadcast by a radio transmitter. It is a way of creating a simple, light-weight, open media system. Recently an orange tent was erected in Vienna's Museumsquarter by the Public Netbase. This tent is a "sonar" media installation, where passers-by and remote users can listen to and interact with TextFM. The messages can be sent to the installation from a mobile phone or via http://basecamp.netbase.org, and may be read in English or German. The installation in Vienna creates a compelling interface between between public and private space.

MK: Matt, I remember when I met you in London, you were talking about the difficulties of implementing the German text-to-speech. How was this overcome?

MF: The problem with German text-to-speech came about because the first version of the software is written in MacPerl. The text-to-speech software that we use in this version is that native to the MacOS. As they only supply English or Spanish text-to-speech, we found that working the project in other languages was effectively blocked.

In order to get a version working German TTS, Ivan at Public Netbase put together a version of the program using MBROLA. The current state of the Linux version is that it's very stripped down compared to the one for the Mac, so some of the more subtle aspects of the program are missing. At the same time though, it benefits immensely from the way that non-proprietary software allows for certain kinds of development away from strictly corporate rationales – in this case, the use of a language which hasn't been globalised.

It's hoped that other language-implementations will be taken up by different groups wanting to use the system in their areas.

MK: I tuned into the stream at http://basecamp.netbase.org/, and sent messages, and something that I enjoyed doing was creating a "conversation" with other users/bots/noises at the time. Are you interested in the possibility of spontaneous narrative?

MF: Sure. One of the things that imporessed us when the system first went live in the Vienna version was how quickly people invented ways of using the system: Sloganising, conversations, insults, meeting arrangements, flyering for DJ sets, asking questions, setting up conversations. For us one of the key things has been to have the system operate as a way of finding out what potential cultures of communication are out there: And people always exceed and confound your expectations in one way or another.

A fundamental impetus to the project is to find out how latent capacities in different media technologies can be found, be mobilised, and then mixed with those of others. But the project also releases latent social capacities and this is what can be found revealing and inventing itself in the texting going on through it.

MK: Does the use of SMS shorthand often create unintelligible messages when read aloud? What are your thoughts on this disconnect between messaging and voice?

MF: Sure. We had to think this one through before establishing the design of the system. Texting short-hands exist in several different kinds of relation to "fully written" text. Acronyms, character substitutions, homonyms and word compressions – missing out on the vowels for instance – are all used in day to day texting. However, in order to be able to get the computer to read them from the phone as they come in and then read them out as the "fully written" version we would have to have a corpus of all these different kinds of short-hand built into the software. Every string of characters would have to have been checked against this list as it came in in order to see if it needed "translation". There were two problems. Firstly, we couldn't possibly hope to keep up with the rate of development of such short-hands. Secondly, some sequences of characters are used in more than one shorthand. The computer would then have had to analyse the semantic context of the string in order to determine which use was intended. Obviously, this would have been a bit difficult.

Instead, we chose to have the system's incapacity to make such interpretations audible up-front. You might use it once assuming it to understand your particular argot. But the next time you won't. That way people can use it without expecting more than they get – and develop new ways of playing with it on that basis. For instance, we've noticed some users are getting the voices to sing or generate simple beat structures by feeding certain mixes of characters in.

MK: Are there strategies in place to use this for subversive or performative purposes? Or will this be unspoken and allowed to flower organically? I could forsee some interesting manifestos, speeches, and revolutionary thoughts being safely broadcast to a chosen receiver location. Also, it's interesting that people have started creating beats and music – performers might choose this as a venue.

GH: I like organic flowers.

MK: When listening to the stream, I am reminded of the lost practise of ventriloquism. Have you received any feedback from your users on the impact of this ability to "throw their voice" from anywhere in the world?

MF: We've demoed the work a couple of times, left the thing running in the background. Usually, most of the crowd have the chance to send a message in whilst this is going on. From this we definately know that the capacity to heckle a speaker with their own set of gadgets is pretty damn inviting!

It'd be good to see the thing used as a Robo-MC for parties and sound-sytems too, get the crowd chatting directly over the music

MK: Are there future plans for other installation-based projects around the world? A multiple-node physical network to compliment the wireless network?

GH: We tried linking Amsterdam and Vienna using two phones. One in each city, then combining the SMS from both places on a single server. I am interested to explore this notion at some future date. Also what might be interesting to you is, I set up a stream of messages from Pres-Bush to the server. This was done by analyzing his speaches since 9/11 then calculating the probability of what he would say next.

Here are some examples.
sent: #?Lev3p3r2 America will not only live, but as it to act.
sent: #?Lev3p3r2 At the state; Respect for her example of thousands
of job should participate in applauding your life we must act as you
very much.
sent: #?Lev3p3r2 I want to help the best that we think we've made
stronger than unemployment checks – will prevail.
sent: #?Lev3p3r2 We were carried forward to stall this conflict and
in case of humanitarian aid or religious tolerance.
sent: #?Lev3p3r2 He's missed his lost loved ones? Under this recession.
sent: #?Lev3p3r2 I love children, and defeated? (Applause.)
sent: #?Lev3p3r2 Americans did not only by day – you aren't sure
good Lord of purchases? (Laughter)? I want to win the fine staff of
the goal – and for Africa?
sent: #?Lev3p3r2 ?? To the world. at
/Users/harwood/perl/markov/mkmarkovlnk.pl line 120.
sent: #?Lev3p3r2 The situation runs for Syria to those things
present nor angels nor depth, can be a grave and the past 21 weeks to
know the United States military.
sent: #?Lev3p3r2 The men and to begin, Nellie, by Congress who have
shown overseas: We'll increase funding to build up the momentum of
this recession.

MK: You have set this system in place and as you put it, "people always exceed and confound your expectations in one way or another". Are there any properties of the system that are remaining unexplored, or that you would love to hear more of?

GH: Making an open server in which streams of SMS could be generated in different countries and added, with maybe some bots adding things as well.