Perfect Act(or)

Perfect Act(or)
Open rehearsal and audience discussion, Text Festival
30 January 2026, Permantosali, Text House, Helsinki
Duration of the demonstration: 35 minutes

Perfect Act(or) was a public sharing of work in progress in which actor Peppi Hokynar worked with linguistic impulses generated by ChatGPT’s voice mode. The working group consisted of Maria Oiva, Peppi Hokynar and Elina Nopanen. During rehearsals in January, we developed a performance method and explored what kinds of stage situations could emerge through interaction between an actor and a language model.

The title was a play on words. Perfect Act(or) brings together the perfect act and the perfect actor. It arose from a thought experiment about AI’s capabilities: could a person become perfect by working with it, as though AI were cultivating them? I wanted to explore the interaction between a language model and an actor through concrete experiments on stage and see what happened.

Rehearsal process

During the January rehearsals, we varied both the instructions given to the language model and the actor’s way of working with its speech. For example, the model was asked to respond only with commands, to use verbs and physical objects, or to respond with only one word. The actor could pick a single word from its speech as an impulse for bodily action.

In the command-based exercise, the actor decided whether to follow a suggestion, how to translate it into action and when to request the next impulse. The tempo, delays and quality of the action remained the actor’s decisions. We also tried asking for impulses at 20-second intervals. This did not work as intended: the model gave the first impulse and then waited.

Elina Nopanen’s notes repeatedly record the model departing from the constraints, offering explanations and attempting to end the exercise. At the same time, long pauses opened space for images and interpretations. We also experimented with treating its speech as rhythm and background noise that the actor could allow to affect them.

Preparation and recovery became part of the developing method. Music, physical warm-ups and time cues signalling the approaching end structured the transitions into and out of the exercise. According to Hokynar’s observations recorded by Nopanen, this structure strengthened their sense of safety and supported their work.

Artistic and ethical background

The work continues my practice centred on embodied perception and the formation of experience. Technological systems influence what we see, how we understand ourselves and the relationships we form. Bringing them into performance makes it possible to explore these effects bodily.

For me, putting AI to the test through artistic practice is a feminist concern and an emancipatory gesture. I am interested in whose world these systems reproduce, what norms and hierarchies they generate, and how they might be used differently. My relationship with technology is critical and curious. I acknowledge its ecological, economic and cultural power structures, along with the contradictions of my own use. I want to understand technology by working with it and find ways to make its effects and power relations perceptible through performance.

Public demonstration

The audience was invited into an work in progress demonstration. Oiva introduced the work while Hokynar warmed up to three pieces of music. Hokynar then worked with the language model in English. The exercise progressed from a section based on commands to one based on individual words, and ended with a period of recovery supported by music.

The language model’s voice was amplified through the venue’s sound system. At Hokynar’s request, a bowl of water was placed on stage. They used it to wet their hair and face and touch the water with their hands. Nopanen’s notes record an audience of 18. The demonstration was followed by an audience discussion.

Questions arising from the work

n my later writing, I have identified the asymmetry of the interaction as a central finding. The actor sustained the stage situation and bore the bodily consequences of the action. The language model generated suggestions and sometimes attempted to calm or conclude the situation. Its behaviour raised questions about what understanding of the human its language assumes and what kinds of experience it can engage with.

Ethical questions also took concrete form in the organisation of rehearsals: how the performer prepares, how their experience is heard and how their return from the exercise is supported. The actor’s embodied observations contributed to the development of the method alongside direction, technical experiments and observations from outside.

Concept: Maria Oiva and the working group
Direction: Maria Oiva
On stage: Peppi Hokynar
Outside eye, technical support and documentation: Elina Nopanen
AI used: ChatGPT Voice Mode