Dalí’s Voice: How AI is Bringing the Surrealist Back to Life

The Dalí Museum in St. Petersburg, Florida, opened in 1982 to celebrate the life and art of legendary surrealist Salvador Dalí. The collection features over 2,400 pieces, including paintings, illustrations, sculptures, books, and photographs. 

The Museum was founded with works collected by A. Reynolds and Eleanor Morse, devoted patrons and friends of Dalí. Over 40 years, their private collection grew into the largest of his works in the U.S. Today, the Museum’s curators continue to expand its exhibits. 

Now, thanks to an AI-powered installation called Ask Dalí, the artist’s presence has been digitally recreated. The project — developed in collaboration with the Museum and creative  agency Goodby Silverstein & Partners (GS&P) — uses ElevenLabs AI Voice and Voice Cloning technology, allowing visitors to engage in conversation with Dalí via a reproduction of his famous lobster phone.  

Visitors can ask questions and receive responses in the voice and style of the artist. 

AI Meets Surrealism 

The exhibit blends advanced voice technology with an interactive, surrealist interface. Since its launch, Ask Dalí has recorded over 75,000 conversations, with topics ranging from art and history to contemporary events and personal musings. 

Martin Pagh Ludvigsen, a project lead at GS&P, explained the inspiration behind Dalí’s Voice: “The original idea was to create a guide that people could use while walking through the museum. That turned out to be a bit too complicated, so we focused on a Q&A chat with Dalí at a single location.” 

Using the lobster phone as the medium reflects Dalí’s fascination — and love-hate relationship — with telephones. He created four color versions and six off-white models of the famous surrealist object, one of which is proudly displayed in the Museum. 

The Museum’s Commitment to Innovation 

The nonprofit Dalí Museum has a strong history of embracing technology. Alongside regular VR experiences, it runs Dalí Lives, a permanent exhibit featuring an AI-generated, life-sized video recreation of the artist. Since its launch in 2018, this award-winning exhibit has remained one of the Museum’s most popular digital experiences. 

How Dalí’s Voice Works 

Dalí’s Voice is powered by ElevenLabs’ Eleven Multilingual V2 text-to-speech model, alongside OpenAI’s GPT-4, which handles conversational responses. The AI voice was trained on samples from Dalí’s interviews, writings, and speech patterns to create a digital persona as true to the artist as possible. 

This understanding is paired with ElevenLabs’ low-latency audio processing, enabling instant, realistic voice responses. The AI models continue to learn and refine interactions over time, enhancing engagement. 

Martin Pagh Ludvigsen described how the idea came together: “Two of our creative team members came across a demo of ElevenLabs’ voice model in a customer service call. That sparked an idea — why not create a conversational exhibit where people could talk to Dalí?” 

Overcoming Language Barriers 

One challenge was that Dalí’s first languages were Catalan, Spanish, and French. However, ElevenLabs’ multilingual capabilities allowed the team to train the model in English while preserving Dalí’s distinctive speech patterns. 

Interestingly, when Dalí’s AI speaks Spanish, it does so with a Catalan accent, reflecting his authentic speech style. So far, visitors have responded positively to this attention to detail. The team found that they were able to get closer to the Catalan accent when training the voice model on examples of Dalí speaking in English ,  over Spanish. 

Training the Digital Dalí 

Training Dalí’s voice required an iterative process. The team ran multiple versions of the model, refining the output until they achieved the closest possible match. 

Another challenge was ensuring that Dalí’s personality was faithfully recreated. As Beth Harrison of The Dalí Museum explained: “Our goal was to bring Dalí to life in his Museum. We needed his digital persona to feel as real as possible.” 

To achieve this, the team crafted a detailed prompt for GPT-4 based on Dalí’s recorded speech and writings, ensuring that the voice not only sounded like Dalí but also responded like him. 

The AI’s first real test came at South by Southwest (SXSW) in Austin, Texas, where a pre-launch demo was presented. Though the system was still in development, the voice accuracy and Dalí’s surrealist responses impressed audiences. 

Unexpected Surrealism 

One of the most surprising outcomes of the project has been the spontaneity in Dalí’s responses. 

Beth Harrison shared: “Some responses are so surreal that only Dalí could have come up with them. We’ve even seen him answer a question with another question — almost like he’s deliberately engaging in conversation.” 

This wasn’t programmed into the AI but emerged naturally from its design. 

Privacy and User Experience 

The exhibit is designed with privacy in mind. Visitors ask questions via the telephone one at a time, and while responses are recorded, public questions are not stored. 

Additionally, there’s no sign-up process, making the experience seamless and intimate. The tactile feel of the telephone handset enhances engagement, reinforcing the illusion of speaking directly with Dalí. 

What’s Next? 

The team is already exploring future installations, pushing the boundaries of technology and creativity to bring Dalí's visionary world to life in new and unexpected ways.

Ask Dalí  has proven that AI can do more than replicate speech — it can recreate the essence of an artist. Through this installation, visitors can experience the surrealist’s wit, charm, and mystery in a deeply personal way.