Creating a Cinematic Opera with Artificial Intelligence

Tjaša Ferme: Hey theatre, science, and innovation fans. This is Tjaša Ferme, your podcast host for Theatre Tech Talks: AI, Science, and Biomedia in Theatre. A podcast produced by HowlRound Theatre Commons, a free and open platform for theatremakers worldwide. Tune in.

I'm talking to Ellen Pearlman today. Ellen Pearlman is a New York-based media artist, curator, writer, and critic. She's a visiting research scholar at the New York University Tandon School of Engineering and a new works artist grantee at Harvestworks in New York City. Ellen, I'm so excited that we're speaking. I first heard about you in around 2017. You were at the time the only person that was really working with EEG headsets on stage and I would love to hear from you a little bit maybe of your own personal story and trajectory of how you even started using that, what you were looking for, and what kind of effect and satisfaction those gave to your creative practice.

Ellen Pearlman: Well, I think it started in 2013 when I was aware that brainwaves were going to be used in human biometric surveillance. I was very clear and I knew that was going to happen. I went to Hong Kong and I started a doctoral program where I was researching about it and the idea of the brainwave opera came to me because the question behind it was: is there a place in human consciousness where surveillance cannot go? During my studies and research, I found out a tremendous amount about what's called the semantic brain, which is not the “left brain, right brain” myth that everyone remembers and thinks of, but showed a very, very clear mapping of how the brain works. It came out of the research of the Gallant Lab in Berkeley and it was the work of Alexander Huth, H-U-T-H. And it showed very clearly how the brain is segmented into different parts of recognition that both understand conceptually and light up if it's being scanned physiologically to categorize in a sort of visual semantics and cognitive semantics intertwined with language object recognition tied to verbal recognition.

And that proved to me that it was the beginning of the age of brain mapping and brain surveillance. And so I began to wonder where is it that surveillance could not go? And I went very deeply into Elon Musk's Neuralink. I went very, very deeply into brain surveillance, fMRIs. I mean, I dove into the Department of Defense. I dove into IARPA, I-A-R-P-A, which is a special wing of the Department of Defense dedicated to biometric surveillance, or it was at the time. And I discovered that not only in the United States, but many countries all had brain initiatives to map the human brain in the same way. I also went into the research of the Japanese who were doing studies on dreaming and the semantic brain, which I found really compelling. And so I asked the question what cannot be mapped in the human brain? And one of the issues, or not issues, one of the things I thought about was faith because I didn't see faith as having a mappable quality.

So then I thought, "Okay, faith." And I was waiting for a story about faith to come to me. And somehow I stumbled upon the story of Noor Inayat Khan, who was a Sufi Muslim princess. And her father, Hazrat Inayat Khan, had brought Sufism to the west and Noor was actually half American. And during World War II, she worked as a secret agent for the Allies in Nazi-occupied France. And she was a wireless operator, which to me said she was a tech girl, and she would string up copper wire in Paris and transmit the movement of the Gestapo back to the allies. And she was caught twice by the Germans and she escaped twice and she was finally executed at Dachau as a political prisoner. And so that was a very compelling story to me because they couldn't break her.

She didn't give up any information. So that was a story of her Sufi faith and I was amazed that there was a Muslim woman, secret agent in World War II, fighting for the allies. I didn't know about this and I thought it was a great story. So that brought together surveillance, consciousness, brain mapping, faith, and a very amazing story, true story.

Tjaša: Wow, super fascinating. Actually, I'm aware of all this research. I actually know of Jack Gallant. He was basically a schoolmate of the neuroscientists that we worked with when we were developing the neuro role model project. And so I remember his Ted Talk about different areas of the brain that hold different notions. And obviously that was amazing, but there was a huge criticism of his work, which was that this study was made only about on seven people. So it was a very, very small group. But also, I mean, they say that basically every individual has their own spots and these are somewhat of areas where these notions live. But obviously this was super compelling and I'm very curious to hear that... Actually, that was ten years ago. About ten years ago. I'm very surprised that Alex Huth, who's now at the Texas University and just created a semantic decoder, that he was working with him already at the time.

Ellen: He was a graduate student. He was his graduate student of the Gallant lab and worked in the Gallatin lab until he moved on.

Tjaša: Yeah, super fascinating. I love it. Small world.

Ellen: Great you know that.

Tjaša: I'm curious why he decided to work with EMOTIV because EMOTIV already basically has these categories of excitement, attention, relaxation, meditation. Sometimes they have more categories, but I remember when my team of neuroscientists worked with that. They were kind of skeptical because they were like, "Well, how did they decide? What's really the science behind it?" So did that bother you? How did you feel about it?

Ellen: First of all, I did very rigorous testing on different headsets that I had access to. Now, of course, I found NeuroSky and Muse to be cute, but not workable because they only worked with alpha and beta waves. I found that Cognionics was really nice with the silver tipped ends, but it was $20,000 a pop, so that was out of my reach. That's what a lot of people were using at the time. And the thing about the EMOTIV, which was very good for a creative performance, was there was about between a 60 percent and 80 percent accuracy rate of emotional valence, which for an artistic performance was just fine. And neuroscientists can criticize me all they want and I'm not really interested because for what I was doing, it was good enough. And the second and most important part of it is it had an OSC out ability called Mind Your OSCs. And that let me pipe it out to MaxMSP. Now, what happened in the ensuing years after that is they discontinued Mind Your OSCs and they made it OSC out.

They in-housed it and then they made it a subscription. And I got a hold of them because they featured me in one of their newsletters. So I actually got a hold of them and I yelled at them and I said, "You have just ruined every artist's creative practice who's ever used your workflow." And I got a hold of the developer who was Vietnamese, who actually made Mind Your OSCs, and I yelled at him and I said, "You have ruined my life. I mean, I can't work in this realm anymore because of what you're doing and also you're making it 64-bit and you're basically taking something that was rather open to the creative community and you are taking it away from them." So he sent me the entire package of Mind Your OSC and said, "Here it is. It's written in C++. You're welcome to adapt it. I give it to you freely." And I said, "Great, like I'm a C++ developer. Thanks, but no thanks."

Yeah, I was like, "Gee, thanks. I love you, too." The other thing is that I also, in 2013, started working with OpenBCI and sponsored their first Kickstarter at CultureHub, La MaMa. And they raised $250,000 for that and they went... So for a few years I was working with OpenBCI, but they had absolutely no interface where you could do anything but measure alpha, beta, theta and data waves, but their chip was $500. Now, OpenBCI has the Galea, which is full stack of everything, including VR and immersive and everything. And it's $20,000, so there goes that particular tool. So in that sense, for the most part brainwaves have been priced out of the artistic market for the most part, unless you have either a corporate sponsorship or a product sponsorship and a team of rigorous engineers. It's just not possible for creative practice to use BCIs anymore because of the commercialization of that space.

Tjaša: I feel like all of your pieces have so many layers. Language Is Leaving Me, I watched the cinematic piece, then I watched the performance with the cinematic piece, and then I watched your interview. And only at your interview I literally put everything together because I was like, "Okay, so which part of it is the opera?" Until I realize, and we'll later on talk about this, but it's basically the muscular sensors are the ones who are basically, according to the biodata, creating a soundscape. Is that correct?

Ellen: Correct, yes. And just so you know, that's only part one of proof of concept. I want to say that when I dropped that AI cinematic piece, that was four months before OpenAI premiered Sora. And so now I am working, I'm actually going to Germany in three weeks to a high tech artist residency where they are going to figure out if they want to develop this. And I already have been working on the tech for it into an actual full-blown opera. So that was just part one, proof of concept of the cinema part.

Tjaša: I mean, okay, let's talk about this. And so what actually happens in the movie is at the beginning you're showing the documentary footage and then on the other side, on the split screen, you start showing basically image depositories from different languages. So you're using translations from English and you're putting them into Yiddish, Chinese, Tamil, and Xhosa. I want to know how this was made. So did you actually went to Latvia and you got the book and you found that story and that happened to you or that was written in the story?

Ellen: That really happened. Yeah.

Tjaša: That happened. Okay, cool. And then you had some documentary footage from your own trip and then you had some historical footage that you were incorporating into a narrative A, which is your narrative and you created. Is that correct?

Ellen: Partially.

Tjaša: Partially.

Ellen: What I had is first I developed during the pandemic, I developed the idea with the Columbia Digital Storytelling Lab and so I had a little proof of concept visual. I was trying to make AI cinema starting in 2020 and in 2021. It was impossible. And then I was actually prototyping in Google Colab and that was pretty horrible, although I did make a seven-second AI video at that point in early 2021. So I just made the movie and then shortly in 2022 LAION 5B came out, which is the five billion visuals. And Automatic1111 came out and Stable Diffusion came out. And that allowed me a new working environment that I had to learn from the ground up and I had to beg and borrow a high-end server in Hong Kong that someone loaned me so I could pipe in with a VPN into Hong Kong to work on rendering the images because I certainly didn't have access to computer with that kind of power.

The second thing is, fortunately, I got a Fulbright in the Department of Mathematics and Informatics at the University of Warsaw and I chose the Department of Mathematics because I wanted to develop this further. And I ran a co-lab and in that co-lab, one of the participants was a cinematographer. And the cinematographer said, "I am going to go on my own dime to Latvia and film that SERDE residency and the town of Aizpute.” He goes, "I need to do this." So he went and when it opens with the blue sky and stuff, it shows the little town of Aizpute, that's his footage. I had substituted stock footage in my original prototype, but he went, "No, it's got to be original footage." And he went out and he filmed the real ones and then of course the rest was archival footage.

Tjaša: Yeah, you even had Nosferatu in there. I was like, "Oh, my God." Why Nosferatu?

We are at one of the largest cultural shifts for the next two hundred years.

Ellen: I'll tell you why I had Nosferatu. A couple of reasons. One is the cultural references because there's a lot of cultural references to early cinema. And that's where we are with AI right now. We're in the age of early AI cinema. And the second was the underlying antisemitism of a certain cultural time, which would have references to those well-versed in film history. When I show it to Asians, let's say, they don't understand the imagery of Nosferatu. It isn't in their film history, so it's culturally specific. But also, AI cinema right now is in the exact same position that the transition from silent cinema to talkies was in and the transition of the agricultural age to the industrial age was in because we are going from the post-atomic to the age of AI and quantum. And so we are at one of the largest cultural shifts for the next two hundred years. And so I wanted to go back to the origins of film and cinematic history and the transition from photography to film. So that's why I put in those references. It wasn't arbitrary.

Tjaša: No, no, no, it sounds really rich and thank you for this clarification. Yeah, I love how you're putting it that we're sort of in a phase of growing pains. And also I want to say we've been playing with different AI softwares that create AI movies before Sora came out and this is painful work. So you working on basically seven second clips and stitching it together and exporting it and modifying it and I don't know what else you were doing to it. This is like, you're dedicated. This is real work. This is really hard work and it's also so frustrating because it's so inconsistent in so many ways.

Ellen: Yes, and that's really exciting.

Tjaša: I love it. You love problems. That excites you. I love it. I would be frustrated.

Ellen: Because when you understand what's under the hood, which I would say I understand what's under the hood both with GPT and with AI cinema, I mean, I'm not a computer scientist at all, but I do understand more or less how it works. You then see the deep flaws. And the point of this, and I don't want to vary off brain waves too much at this moment, but the point of this is that the inherent flaws, especially in the development cycle of the big three companies, meaning OpenAI, Microsoft, and Google, and Meta, that makes four actually, are being glossed over faster than the speed of light. And the flaws and the implications are very serious. And especially in North America, nobody wants to deal with that at all. They just want more, better, faster, more intense. Here we go, commercialization, billion dollars, trillion dollar industry.

Tjaša: Let's talk about those flaws. What are some of the most detrimental flaws and what are the consequences we're looking at?

Ellen: One of the reasons in the way I was making AI cinema, first I was working with VQGAN before Stable Diffusion came out because you couldn't work with DALL-E, which was a diffusion model because it was locked up. And what I found is that linguistically most cultures, both with their language and with their image repositories, were not represented at all. And I did experiments in 2021 in Google Colab, which were multilingual, and discovered that the misrepresentation and mis-categorization was through the roof. Now, when stable diffusion came out and LAION 5B, I then went very deeply into what is called the aesthetic scoring. There is something called aesthetic scoring of stable diffusion. And I went into what it was, where the repositories came from, how the scoring was determined, and who scored it. And that’s when I started getting very alarmed.

Then I went into the types of styles that were available for visual representations of Stable Diffusion and I became even more alarmed because there was a sort of... This is a terrible term, but I’m going to use it, whitewashing of images. And I don’t just mean racially, although that’s certainly true. I also mean aesthetically and conceptually. So I got really alarmed and Language Is Leaving Me shows actually how one representation, one I like to use because it’s so graphic, is in the story there’s a line that says, “And a woman held an infant to her breast. It wouldn’t let go. She wouldn’t let go.” And there’s an old archival scene of two little children grabbing a woman who’s sitting on a step. It looked like 1930s Germany. And when I translated that into the Mandarin language and I ran it through Stable Diffusion in Mandarin, image to image and word to word it brought back softcore Chinese porn. I didn’t do that.

Tjaša: God. How? Jesus Christ.

Ellen: Well, no, I understand how absolutely. How it did that, it's very easy to understand in layman's terms. There are two ways. One is every second of every day there are web crawlers going around the entire world, sucking up images every day from the internet. So it's contemporary. The second thing is the majority in the Chinese language, which override any other content, are girly, massage, softcore porn sites.

Tjaša: Wow.

Ellen: And so what happens is the repositories were full of that and the thing it brought forth, so when you would do the word breast and you translate that to Chinese and... Now, if you know in the original, there's a clothed woman with two clothed children who are just hugging her around the neck. In the Chinese version, there's a almost naked Chinese woman with exposed breasts who's very young and very voluptuous. And that's how the Chinese interpretation of that phrase hierarchied in aesthetic weights.

And the children, the babies, I said “she held an infant,” the babies are gone. When I do that in, for instance, Yiddish, there is no woman and there is no baby. It's just a man. When I take that same phrase in Tamal, it's a man with a beard in a sari standing next to another man. So I started seeing these absurd inconsistencies and that was incredibly compelling to me. Absolutely it gave me... I mean, then I started seeing, I think the current term for it is, the Loab, L-O-A-B. I think that's the term for it, or the psychic unconscious and dark side of AI rendering. And then I got really excited.

Tjaša: Yeah, okay.

Ellen: So that's why that movie is a little hard to understand initially because I am doing comparisons in multi-languages of image repositories. And so I know it's a little bit hard to understand, but when it gets made into an actual opera, it will be easier. This was just proof of concept to make the movies, actually.

Tjaša: No, no, no, but when you hear the explanation, it makes a lot of sense. It's just, like I said, it's a lot of layers. So you have to figure out which layers are, how they're layering, what's going together and what's separate.

Ellen: But it's also, it's not just language translation. It's both language, it's text to image, and it's also image to image.

Tjaša: See, more layers. See, that's more layers. How do you layer that?

Ellen: Well, it's using computer vision and algorithmic recognition. In other words, in all these algorithms that we use in visual AI, there is some really simple computer vision going on. And what I mean by simple is simple to grasp, not simple to make it happen. And some of that is edge detection and shape detection. I mean, I'm looking at you right now and you're looking at me and what do we see? We see a round circle for a face and then a shoulder. I mean, the line sketching, that's what image to image compares. That's a human and that's how object recognition works. And that's what computer vision is, how to recognize the object. So under the hood in all of that, when you begin to see the outlines of things and computer vision is going back and forth super fast to say, "Okay, that square equals this square, that circle equals this circle, that vector arc equals this vector arc." And that's image to image comparison. So I'm doing both linguistic comparison in multi-languages and image to image at the same time.

Tjaša: And would you say that one of them is primary, that text is the driver and maybe the image recognition is something that dictates the synchronization of rhythm between let's say the two languages?

Ellen: Okay, that's something that I'm exploring very deeply in the new piece I'm making. And I would say that that is what is called weighted. And what I mean by weighted is if you put it on a scale, you can balance it different ways. So using different, what's called, checkpoints, different seeds, different filters, that can be changed. And I'm working very, very deeply on the new piece to come up with an aesthetic that I like as opposed to an aesthetic that is commercially served up. Because I'm, again, trying to work outside the constraints of normal commercial apps. And in order to do that, I need access to very high-end computers that are water cooled they're so high-end. And in order to do that, it's not easy to get my hands on those because I can't buy them. So I have to, again, beg and borrow them all around the world. And that is literally what I'm doing because I'm not the big companies and I don't have access to those kinds of things.

Tjaša: But you've been resourceful. You've been really successful in finding ways to make this work and to continue the research and, like you say, do deep research, which is really, really time-consuming. But I think that the seed for all of this lies within your personality, which is like you love what makes a lot of people frustrated, what's difficult. You thrive on what's difficult, it seems a little bit.

Ellen: Thank you for that, but I would rather say that I ask difficult questions because I have a very deep understanding of the misuses of systems of categorization. And the potentials for misuse of these systems of categorization are being constructed now. When I said we're in the transition of this age, this is the moment where these systems are being constructed. And once they're constructed, it's almost impossible to undo them. It's like saying you're laying down the major highway arteries of an entire nation. And once you lay those roads down, you're not going to rip the roads up and change their path. It's just not going to happen. So we are currently at that moment in time because right now everything is through, let's say, large computers and servers, but eventually it's going to shift over.

And I am absolutely certain of this into quantumized vectors of information flow. Which are going to happen so fast and the repositories are going to be so strong that mistakes are going to be impossible to unwind. And resources in order to challenge that, are going to be very difficult because it's not going to be something the general public is really either exposed to, aware of, or knows how to cope with.

Tjaša: It's almost like we're building in values, so to speak, and it's a similar problem, that let's say other large language models had, like Lambda and whatnot. Once the biases are in the system, you can't really rip up the system and start from scratch. So whenever there is a bias, the bias will stay within the system unresolved. So while we're still in the young age or early age, it's time for us to sort of be the messengers of what's not working and what's potentially causing harm and misrepresentation so that there's a potential of still undoing and improving basically the bottom layers of bricks, right?

It's not just image representation. It's jurisprudence, it's healthcare, it's banking, it's real estate, it's migration. I mean, these are the systems which are the conduits of human congress.

Ellen: In a simple way, yes. But the thing is, the implications are much more profound. They're profound in the sense of it's not just image representation. It's jurisprudence, it's healthcare, it's banking, it's real estate, it's migration. I mean, these are the systems which are the conduits of human congress. And these whole systems of human congress are going to be shaped into the digitized nations in the non-digitized nations. There's seven thousand languages in the world. Probably five to eight of them are going to be hierarchied and privileged in making decisions and holding repositories of memories for all the rest of them. And that's what I see. I'm not dystopic, but that's what I really do see. And the implications for that are seismic.

Tjaša: I'm guessing that your choice of the languages that you chose for Language Is LeavingMe are not random. So you have Yiddish, Chinese, Tamil, and Xhosa. Why did you choose these and what did you learn from each specific one? You gave us some examples already, but how deep does it run?

Ellen: Although my family came here in 1906 from the Pale of Settlement of Eastern Europe, I didn't know until about four years ago that they came because they were escaping the pogroms of 1905. I didn't even know that there were pogroms in 1905 because nobody talked about it ever, ever. So that's the personal because I can't talk about anybody else's issues. I have no right to, so I went, "Okay, I've got plenty. I'll take some of those." And so I don't know Yiddish. I live in New York. There are Yiddish words thrown around in the parlance of New York, but I don't know Yiddish. So that's the reason I chose that, also because of the cursive script. And then I went, "I've been to China a lot," and I went, "There's another cursive script, Mandarin." And then I went, "Okay, Tamil." And then I chose Xhosa because I wanted—even though, Xhosa isn't a cursive script, it's a South African language in English, in normal alphabetical letters. And I said, "So let's get a wide net of places in the world."

That's why I use those scripts, because it's not necessarily that they're all cultures of diaspora, but what they are is they're not well-built repositories. Political campaigns and movie posters comprise most of the Tamil data bank. Most of the Chinese data banks are either, like I said, softcore porn, calligraphy. Everything is softened down and flowered out. There's no allowance of harsh imagery, except for the softcore porn. In Yiddish, it's super male-dominated, unbelievably male dominated. I mean, you say the word woman and a man shows up. That's what I mean. That's what I'm trying to say. It's that dominated, so it shows the cultural failures of repositories at this moment.

Although there'll be now LAION 12B is supposed to be coming out, which is twelve billion imageries, but they've discovered child porn in LAION 5B and they had to really deal with that. Yeah, because it's scraping the internet. They took it out. It's complicated is what I'm saying, but that's why I used those languages. And it was fun for me to use Yiddish because I went, "Okay, so what's this going to show?" And it was shocking to me. I was shocked by all of the languages I used, all of them, what they came up with.

Tjaša: Yeah, it feels like it's not only showing the deficits of AI, but it's also pointing at some flaws within cultures and misrepresentations of cultures, right?

Ellen: Well, let's say you are working with a nomadic culture and there are many cultures. Let's just say the nomads of Siberia. They're basically not a written culture. The reindeer herders, let's say, of Siberia. They are an oral culture. They don't have archives. They are not digitized. So when you begin to represent their culture, you start probably with the movie Nanook of the North, which is a very old anthropological movie and that has received a lot of blowback of being colonially representative of the culture that was documenting it. But that's about as early as an archive in imagery as you get. And there's not a lot and it's all done by documentary filmmakers or archeologists. It's not done by the community themselves.

They're not archiving their life and they didn't and they haven't until maybe ten years ago or fifteen or twenty years ago. So when you're representing that culture and they're representing themselves, where are the materials coming from to put that into the repository? And who's paying for the digitization of those materials? So immediately you get cultures and societies who are digitally rich and have technological and financial resources to digitize. So in that sense, I used the imagery both of the history of film in Language is Leaving Me, but I also used a lot of World War II footage, free open source, because I'm very aware of copyright infringement. So copyright free, open source visual materials that were basically shot by the US government or the Allied forces at a certain time in history. And that's what I had access to that was free and unencumbered. And that is, even though it was post-war, it was a privileged position.

Tjaša: Let's move on to using EMG sensors. What else can you tell us about them?

Ellen: One of the people on my team was a young neuroscience student and the neuroscience student said, "If you smile, your cheekbone goes up. And if you frown, your eyebrow goes down." And I went, "Great," because it's very hard now, as I said, to use brainwave sensors because of the corporatization of that industry. So I wanted something to the equivalent of what you could say quick and dirty. So first we said EMG sensors and one of the other students said, "I can design a computer board," because I'm in a school of math with computer kids. So this young man designed a computer board and sent it to Shenzhen, China. We got it back. It didn't quite work, so it had to be redesigned and reordered in Shenzhen. It was incredibly cheap, like twenty dollars or thirty dollars, and sent it back. And then I had a sonic artist student who worked with a neuroscience student, because I ran a co-lab, and a cinematographer student. And we were all working together to put this together.

So yes, I had the idea. I knew because I'm very aware of biometrics, how it would work. I supervised them, but they did the individual parts. Which is, as you well know, you have to work in teams for these.

Tjaša: Absolutely. Absolutely.

Ellen: You can't do it alone.

Tjaša: Yeah, yeah, yeah. It's all teams. None of this stuff can be done alone. Nothing in theatre you can do alone. Originally, I'm Slovenian and there's a very famous Slovenian actor that in one of his experiments he wanted to prove that one single person, an actor, can do everything themselves. They can write, they can direct, they can act, they can be their own audience. And you know what? He died during this experiment. So yes, absolutely. When I say that you built it, I mean, you as a team for the specific needs of your project. It means that you didn't go out and just buy something from the internet, so I'm just curious to hear more about what you made as a team.

Ellen: One of the other things I learned very clearly and studied very deeply is human computer interaction, both from the computer side and the human side. And I worked starting with Noor in setting up feedback loops between the audience and the performers. And in order to do that, I had to understand how is a feedback loop created? It's created with presence for the most part. And presence is gaze, touch, movement, voice. And that changes your brain waves. Right now, you're sort of focused in attention and that's brain waves. You know these things very well, but it's the combination of all of it. In order to train the performers for both Noor and Abo, it took over five months. And the reason was it is impossible to fake a brain wave, even if you try. In acting, you can fake an emotion or you can go into an emotion, but in brain waves, your brain waves can't lie.

And that's the issue about surveillance also. So they had to be trained in contact improvisation, which is a dance technique, and they had to be trained how to bring their energy into their bodies. And they had to be trained how to move with that energy because what I discovered is that each performer tended towards certain emotional proclivities. The first performer became angry a lot, normally. The next performer got excited too much. And what would happen is these emotions would override the other emotions. So during the whole performance, they would only trigger one emotion. So they had to be trained to allow the other emotions to arise. Not to suppress their emotion, but to work with it. And at different times in each performance, I would cue them and we had a signal where they would just stop and we would go almost into a meditation because I had to cool them down.

And then the meditation brainwave would emerge and then we made that part of the performance. And that was also so they could reset their brains. Another thing I found is that one performer could only handle a headset on her brain for twenty minutes and another performer could handle it for forty-five minutes because of the Bluetooth and the pressure. So I found there was tremendous variations between performers in both emotional valence and physical capacity and each one was unique. And that took literally, I would say, an average of five months of training to be able for each one to allow who they were to emerge so it would appear seamless. They would walk among the audience, they would touch the audience, they would look at the audience and it all seemed very natural, but they needed tremendous training to do it.

Tjaša: Super interesting how you originate in the reality and the conundrums and specificities of reality, how you don't impose just some abstract command of what you want. And it's really interesting when you're pointing out that we just have some natural proclivities, that everybody's base is different. And you were looking for a clean flute, a clear channel so that you could create a range. And the base was a little bit hollower maybe. Hollow bone. What we call Shamanism hollow bone.

Ellen: Yeah. Well, the head of the School of Contact Improvisation in Hong Kong came in and trained the performer the first time. And I was there when they trained the performer. So when I was in Estonia for the second brainwave opera, I knew the techniques to use. And also, the other thing I would do as we were training is I would let the performers watch their brainwaves live time, so they could see themselves how different actions would modulate their brainwaves. So that was another type of feedback loop where they could self-train.

Tjaša: As a performer, it's so rare to have an opportunity like that for somebody to dedicate so much of time, energy, and method to the process so that not only you can achieve something as an achievement, external achievement of an artistic form, but something internal, that an internal change occurs.

Ellen: Yeah. And that was for them, I'd say it was a deep internal experience. And they learned that they tended towards certain emotional valances much more than they realized as a personality type, even by measuring just four emotions.

Tjaša: Super fascinating. I love it. Thank you so much for this amazing interview.

This podcast is produced as a contribution to HowlRound Theatre Commons. You can find more episodes of the show and other HowlRound shows wherever you find podcasts. If you love this podcast, I sure hope you did. Post a rating and write a review on those platforms. This helps other people find us. If you're looking for more progressive and disruptive content, visit howlround.com. Thanks for listening and have an amazing day.