OpenAI at Lollapalooza Brasil

OpenAI arrived at Lollapalooza Brasil 2026 as the festival’s official AI guide. The campaign was reported as the company’s first marketing activation in Brazil, with a custom ChatGPT for planning the festival and a physical experience inside Lolla Lounge.

That space had two ways into the technology. Two photobooths turned portraits into stylised four-image prints. A red public telephone connected visitors to a voice assistant that could answer questions about the festival.

Bolha created and developed the installations with Fuse and OpenAI, then handled their production and operation on site. The physical language was deliberately familiar: take a photo, pick up a phone, turn a dial. People could understand the starting point before the AI did anything.

I worked on the project through Bolha as a Creative Tech Designer, with responsibilities across both physical experiences.

For the photobooths, I worked directly with the OpenAI team on regionalisation and the local implementation of the software they were developing. I connected that image experience to the equipment we had on site, bringing the camera setup, capture WebUI and printer into one short, legible flow.

For the phone booth, I worked on the interaction, conversational behaviour and the tools that kept its answers tied to the live state of the event. The Bolha team developed the physical and technical hardware; my focus was how the assistant listened, responded and used current festival information during the conversation.

The photobooth experience ended with something tangible. A visitor stepped in front of the camera, followed a short WebUI flow and received a stylised grid of four images, printed a few moments later.

I worked across the handoffs in that sequence. Regionalisation was one part of it; adapting the software to the local camera and print setup was another. Each stage needed to pass the right state forward: a valid capture, a completed transformation and a print job the operator could understand.

A festival queue gives those details real weight. The interface had to be quick without feeling abrupt, and failures needed to be visible to the team before they became confusing to the next person waiting.

The phone booth started with an ordinary action: lift the receiver and talk. On the other end was a conversational festival assistant that knew the artists, the schedule and its own position inside the event.

Time changed the answer. The assistant could recommend a show that was still ahead, recognise that another one had already finished and discuss updates added during the day. Location mattered too. Because it knew where the booth was, it could give directions and talk about distance from that point rather than offering generic navigation.

A static festival FAQ would have aged by the hour. We designed the conversation around information that continued to move with the event.

I helped shape the interaction model and the tools the assistant used to query festival information. Its replies were grounded in a changing event dataset rather than a fixed block of background text.

The current time divided the schedule into what had happened, what was happening and what was still possible. The booth’s location gave spatial questions a reliable origin. Updates made during the festival could change recommendations or add context about a completed show without rebuilding the whole conversation.

This also affected the writing. Over a telephone, responses need to be short enough to follow by ear and specific enough to act on. Regionalisation covered the language and tone, but also the practical way people ask about stages, artists, timing and where to go next in the middle of a Brazilian music festival.

This project compressed a lot of creative-technology work into two simple gestures. One person pressed a button for a photo. Another picked up a telephone. Behind each action sat software, local hardware, operational tools and a live event context.

Most of my work happened where one part met another. I connected OpenAI’s software to the equipment in the room, shaped how the conversation accessed current information and helped the prototype hold up under live festival conditions.

That is what made the project useful to me. The AI was important, but the experience depended on everything around it being designed as part of the same system.