Yuval Sharon has been thinking a lot lately about artificial intelligence and its effect on opera. “Most of the opera world thinks that AI is going to help us with ticket sales and marketing copy,” notes the artistic director of Detroit Opera, known well for his unconventional stagings of inherited repertoire. “But what about the art?” he continues. “Opera is always responsive to technological innovations — how can we pretend that AI isn’t going to encroach on the art we create?”

Sharon is himself in the midst of devising his own “encroachment.” His new production of Così fan tutte, opening in Detroit in May 2025, will recast the work’s four young lovers as robots, with the old philosopher Don Alfonso and the maid Despina as their programmers. Even more audaciously, Sharon is flirting with the idea of replacing Mozart’s own recitatives with AI-generated adaptations.

This would be the first time a major American opera company will have performed AI-generated music, but the rest of the music industry is already seeing an influx of AI-influenced tracks. The record label Warner Music Group signed its first AI pop artist in September, called “Noonoouri.” “Heart on My Sleeve” is an uncanny AI-generated duet between facsimiles of the pop stars Drake and the Weeknd that stirred deep questions about copyright and using an artist’s likeness and style. In the opera world, “I could feed all of Maria Callas’ recordings into AI, then ask it to create a performance of a piece of music that Callas never sang, and it would sound good!” says Douglas McLennan, the editor of the popular digest Arts Journal. McLennan has taken a keen interest in exploring artificial intelligence’s potential ramifications for the arts.

“That changes the whole notion of ownership of the human voice,” he continues. “AI is going to disrupt pretty much every business model and creative model we have.”

First Drafts

Such musical creations are already within the bounds of what’s possible in the opera world, though no companies are yet taking the technology into such ethically murky territory. Still, the industry is already putting artificial intelligence into use. Administrators are using language generators like ChatGPT to assist in devising marketing copy and grant proposals. Designers and directors are using text-to-image programs like Dream Studio and Ideogram to kick around production ideas.

Take the experience of E. Loren Meeker, general and artistic director of Opera San Antonio, who decided to test ChatGPT’s capabilities on a lark this past summer to help her work through a stack of grant applications. Meeker’s prompt included language she typically uses to describe her company’s work and its engagement with the community. The experiment was a qualified success. Meeker notes that the bot wasn’t very accurate with titles and dates, but that it helped provide a range of ideas to work from. “You can put in specific prompts and ask for, say, five versions of an answer, and then you can pick and choose the language that works best for you,” she says.

“The one thing ChatGPT doesn’t do is ‘perfect,’” says Nicole Malcolm, director of marketing and communications at Pacific Opera Victoria (POV). “The sentences are all correct, yet something feels off. The heart is missing. The copy is always going to need that human touch.” Malcolm first used ChatGPT to help write POV’s 2023–2024 season brochure. The company had commissioned a series of food-themed images to brand its three-opera season, and the brochure copy had to match. Malcolm found synopses of the works in question, fed them into ChatGPT with the instruction to give the copy a culinary slant, and then finessed those dozens of iterations with her team. Aside from copywriting, she has also used ChatGPT as a scheduling assistant during especially busy periods and even to help her outline the structure of a marketing workshop she ran at the Canadian College of the Performing Arts.

Visual Aesthetics

Some companies have started experimenting with AI’s ability to deliver imagery as well as text. Fernando Menendez, marketing director for Opera Idaho, used the image-generating program Gencraft this past holiday season to create a graphic for a Christmas Carol- themed concert. The task was to find a picture that suggested the Victorian era without pilfering an actual illustration for the Dickens novella. Gencraft’s result suggested the holiday and the Victorian era, and Menendez used it, with the concert’s name emblazoned across, in a mailer and an e-blast. “It was generic enough to make a good background,” he explains.

But just as with today’s AI text programs, the current crop of image-generation tools may well fall short of creating graphics that will work in more prominent marketing contexts. When OPERA San Antonio scheduled an extra concert after its season graphics had already been designed, Meeker turned to the online program Canva to see if it could deliver an image in the same style. “It didn’t give me what I wanted, mostly because with the graphics we had a human eye involved,” she says. The company turned to the original designer for the extra image.

Would using an AI image inspired by the artist’s work have constituted a copyright violation? The answer isn’t clear — the legality of using AI to augment or replace creative work is an ongoing conversation in legal circles. “It would depend on a lot more facts and the images themselves,” says Christian Ehret, an attorney at the Webb Law Firm. He explains that ideas are not “copyrightable,” and that it’s difficult to determine whether the definition of “style” overlaps with an idea. “At the end of the day, you really need to look at both works side-by-side... It’s a difficult and seemingly subjective part of the copyright process,” he says.

Age of Anxiety

It seems clearer that creatives using AI tools as a part of their own processes is considered more acceptable than administrators using AI to create or embellish creative work. Some production designers around the country are experimenting with using text-to-image programs like Dream Studio and Ideogram to build concepts with AI rather than sketching them by hand. “Let’s say a director is planning a production of Civil War-era Figaro,” says Laura Lee Everett, executive director of the United States Institute for Theatre Technology. “Maybe he says, ‘I want multiple levels, but not too many stairs. I need six doors and a window to the right onto a garden space.’ Rather than sketch it out on a cocktail napkin, the designer can input the information, and poof! — the program will deliver it,” Everett says. “Then if the director says, ‘Let’s see it in midcentury modern,’ the designer can make the change in an instant.” Once the basic concept is locked, designers can then take it to the next stage by hand.

Using AI in this way has caused occasional backlash online, particularly from creatives worried that this could increase the fragility of the creative economy. At Mission Opera in Los Angeles, California, director Christian Catena and the company’s general director, Joshua Wentz, had roughed out preliminary costume designs for Madama Butterfly using AI. Those images were inadvertently posted to social media and earned some sharp rebukes on X. (“Don’t know where they think they’re getting the budget for a prosthetic third foot for ‘C-Ciu-San’ but dream big!” one user tweeted in response to an AI-generated sketch, later mocking an image where a baby appeared in a Dodgers jersey.)

“We weren’t doing anything illegal or unethical,” Wentz says. “We weren’t replacing people’s jobs. AI was useful to us, and it will continue to be useful to us, but I don’t think we’ll be posting these things on social media until it becomes less of a hot topic.”

Nevertheless, there are panels springing up around the world to discuss the ethics of AI, and some U.S. opera companies are creating AI policies to clarify the ethics behind the use of the technology. LA Opera recently issued a set of guidelines for its employees and contractors. “Just saying, ‘It’s not fair, so let’s not do it all’ wasn’t a realistic option,” says Diane Bergman, the company’s VP of marketing, communications, and technology services. “But we were seeing our employees start to use it, and it was causing some angst.” The company created a six-page document that includes directives like “personnel are responsible for using AI in a productive, ethical, and lawful manner,” and “AI tools or content may not substitute for your professional judgment or ethical obligations.”

Of particular concern to the company was AI’s potential to violate the confidentiality of proprietary information. “Let’s say you want to generate a report, so you cut and paste a list of donor names and enter it in,” says Bergman. “Now that donor information no longer belongs to you: it belongs to whatever platform you use. The problem is that when people use these tools, it’s just them and their computer, and it feels like it’s private. But it’s not. It’s even more risky and more powerful than putting these things into social media.”

The Mission Opera dustup is also symptomatic of a widespread social anxiety over the possibility that AI will replace human employees and copyright concerns about using AI for images. In fact, an Ernst & Young poll reveals that a full 65% of American workers are afraid that AI will replace them on the job.

Still, the technology is not currently poised to replace workers in the opera industry. “We have more on our daily to-do list than we have time to do it,” says Jason Hardy, general director and CEO of Knoxville Opera. “So whether AI saves us five minutes or two hours, it’s just going to make us more efficient. From a development standpoint, that’s one more person I can make a phone call to.” But even if no opera-company employee is currently at risk of being replaced, a command of AI tools will probably increasingly be part of the required skill set for industry professionals, just as in the 1990s internet fluency became a necessity in the American workplace. “Is AI going to take your job? Not tomorrow,” says USITT’s Everett. “But people who know how to use AI will replace those that don’t in the long run.”