Opera Verace Foundation

What happens when you apply serious data science to an art form that has never had it? The Stanford AI in Performing Arts Lab in collaboration with OVF is where the research underlying many OVF programs lives. Its three core areas: AI and data science applied to business data and artist training in the performing arts; AI tools for singer development and casting; and conversational AI for opera education and engagement. The lab operates at the intersection of computer science, musicology, and cultural analytics, a combination that does not have a natural home in any single academic department. It draws on 25 years of methods developed in biomedical data science, including ontology engineering, longitudinal cohort analysis, and machine learning for outcome prediction, applied to a new domain where the data is messier, the labels are subjective, and the stakes are an art form’s survival. #Opera #ArtificialIntelligence #PerformingArts

Building the infrastructure for a new kind of opera presenter Opera Verace, the presenting arm of Opera Verace Foundation, presents opera not as an end in itself but as the mechanism through which breakthrough-ready singers get professional performance opportunities with orchestra, in front of live audiences, with professional recordings that the artists can use for professional purposes. Our upcoming productions and partnerships: a semi-staged performance of highlights from Don Giovanni, in costume, at the Mozartsaal of the Wiener Konzerthaus, conducted by Alma Deutscher with the Wiener Kammerorchester in November 2026, as well as three operas at the Wiener Konzerthaus with Alma Deutscher conducting in 2027; a partnership with the Harvard Radcliffe Orchestra and Boston Youth Symphony Orchestras in which we cast emerging young singers in annual operas with recordings; partnerships with Teatro Nuovo in New York and with Opera Fuoco in Paris to record their productions featuring emerging young singers; and a season we are planning at the MuTh hall in Vienna, with performances and recordings that give leading roles to emerging young singers. The common thread through all of these productions is high-quality audio and video recordings with demo reels provided to the artists for their professional development and for public broadcast to entice new audiences to attend opera. Our partnership model is directed at casting outstanding emerging young singers and funding recordings of their performances, whether in our own productions or those of partner organizations. The result is expanded opportunities to grow the visibility of talented emerging young singers, opportunities that would otherwise be limited. If you lead an orchestra or venue interested in this kind of partnership, we are interested in hearing from you. #Opera #ClassicalMusic #LiveMusic

The problem is not that there are too few talented opera singers available to take on major roles. The problem is that many emerging singers lack sufficient visibility to receive an audition or be engaged by opera houses. With the large number of talented artists graduating from young artists programs each year, video is increasingly important to singers entering freelance careers. A singer’s career today is built as much online as in the hall. Casting directors, intendants, and artist managers often look at video before inviting singers for auditions. Professional-quality recordings are expensive, particularly given the need to re-record repertoire to keep the videos updated. Most emerging singers cannot afford to pay for recording, and recordings at home with an iPhone are usually poor quality. When emerging singers perform, few opera houses record their performances, and even if they do, it’s rare for the singers to have permission to use video footage. This is an important contributing factor to why so few of the many outstanding singers who complete training ever have a career. At present, to be seen and heard, singers must spend large amounts of money to fly to auditions in the USA and Europe, and even invitations to audition may hinge on having video. We aim to improve the access to video and visibility of singers who seek opera engagements. The Singer Exposure and Engagement Network (SEEN) is Opera Verace Foundation’s artist visibility program. We are starting a program to rent a recording studio or hall and provide an AV team to make studio recordings of selected artists. We screen and identify breakthrough-ready singers, present them in live concerts or operas in the USA and Europe or partner with opera organizations who engage these singers, and produce professional-grade recordings of their performances. Each singer retains perpetual, royalty-free rights to use their footage for promotion. OVF obtains a non-commercial license to broadcast the videos, with artist approval, to enhance artist visibility and to raise public interest in opera. The recording rights model is a deliberate policy position: the recordings are critical to help artists emerging from training to be seen by the opera professionals to give them a chance at advancing their career. An enormous investment is made by institutions in training future opera singers, yet there is scant support for these singers after training until they are established. We believe providing leading-role performance opportunities with video rights to the artists is crucial to giving emerging singers a chance for a career, and thus leading-role performances and video are the core of our artist support programs. If you lead an orchestra, opera house, festival, or venue and want to explore a partnership that gives your emerging young singers professional recordings, we would like to connect. #Opera #EmergingArtists #ArtistDevelopment

Data in opera houses can be very messy. We are building methods and a pipeline to fix that. Most opera houses have the same data quality problems: null concept IDs, duplicate patron records, wrong-domain foreign keys, missing fields, and inconsistent coding practices accumulated over years of staff turnover. Before any analytics can happen, the data must be “cleaned” to resolve such data issues. That cleaning currently falls to whoever happens to have the technical capacity, which is time-consuming, costly, or may not be done at all due to resource constraints. To the extent that data cleaning practices can be specified in the form of rules, many data cleaning tasks can be automated, and more challenging cases can be selectively surfaced to human review, reducing the complexity and enhancing the quality of the data. OVF is developing an automated data cleaning and quality scoring pipeline specifically for Tessitura-sourced opera data. It runs concept-decode validation, duplicate detection, cross-source reconciliation, and completeness scoring mapped to POMOP tables. It produces a data quality report a non-technical administrator can read and act on. We are looking for opera houses interested in partnering with us to evaluate and potentially adopt this technology. #Opera #DataQuality #ArtsManagement

Opera Verace Foundation is building a large database of opera singers derived from publicly accessible data (not from proprietary databases) to study longitudinal career trajectories. Such data could be used to understand the factors that relate to a successful career, and could be useful for singer coaching, training, and evaluation. Ultimately such data could inform casting decisions, selecting which singers among the thousands finishing training that should be prioritized for engagement in performances and receive support. We are beginning to study this rigorously through data we record in controlled conditions. #Opera #Singers #CareerDevelopment

One of the hardest problems with assessing the voice in recordings of singers is assessing voice size and projection. A singer can be technically accomplished, musically intelligent, and charismatic on video and still not have a voice that will carry in a 2,000-seat house. That assessment requires an in-person audition, and always will. However, when doing preliminary screening to select worthy singers to audition, it would be helpful to have an idea of vocal size and projection from recordings. Such work is not meant to replace the live audition, which can never be replaced since evaluating voice quality involves many factors best assessed by listeners. But quantitative assessment could be useful in training and initial screening. Opera Verace Foundation is researching AI-based acoustic analysis for singer evaluation and feedback that uses the singer’s formant, singing power ratio, harmonic richness, and low-frequency foundation to estimate operatic projection from recordings made under controlled conditions. We are also studying the limits of this approach: room acoustics, microphone placement, and recording chain all introduce variance that acoustic models have to account for. We are happy to engage with research collaborators. #Opera #VoiceScience #AI

The opera field has no shared language for data. Having a lingua franca of the opera domain is important to look across the opera ecosystem to understand commonalities and differences in what they do and what is effective in tackling their challenges. This can be done using multi-institutional analytics and data mining. Before any meaningful cross-institutional analytics can happen in opera, there must be an agreement on what things mean. What is a “patron”? What counts as an “attendance event”? How do you represent a singer’s career trajectory, a production’s creative team, a house’s programming history, in a way that is consistent across organizations? We are developing the Performing Arts Ontology for Multi-Organizational Programs (POMOP), an ontology for the performing arts, specifically opera but extensible to other performing arts. “Ontology” as used in data science is a formalism that describes standard names for entities in the domain of discourse (for example, donors, productions, performances, singers, ticket purchases), attributes (properties) of those entities, and relationships among the entities. Ontologies enable software and humans to share a common vocabulary and knowledge of the domain in a computable format that enables multi-institutional analytics, data mining, and learning analytical models. POMOP is designed to be extended by any arts organization for their own local terminology. POMOP powers STAGE, and it will be published as an open, community-extensible resource with a portal for browsing and submitting extensions. A scheduled autonomous agent will do initial review of submitted extensions for patterns worth promoting into the ontology. We solicit feedback and collaboration from the community in making this a valuable resource to that portion of the opera community interested in leveraging its benefits. #Opera #Ontology #PerformingArts

Opera companies spend a substantial amount on marketing to acquire or renew audiences, subscribers, and donors. It is very costly to inform decisions using their historical data, since doing such analyses requires tedious access to the data and working in spreadsheets. It can be particularly difficult to easily monitor changes in audience behavior over time after introducing new marketing campaigns or other strategies. The question they need answered is not “who came?” but “who was likely to come back” or “who didn’t return and what factors contributed?” and “what could have changed that?” Aubrey Bergauer contends that understanding audience behavior is the critical need, not simply looking at audience numbers on a given day. We believe that to achieve greater insight into audience behavior, mining abundant data opera houses acquire on audiences (ticket sales, subscriptions, donations, and changes in those over time) as well as data at other institutions and venues is vital to understanding audience behavior and being able to design approaches to attract them to opera. To do that efficiently, a data warehouse that amasses data from multiple institutions is required, or data federation if sharing data is not possible. Such methods have been extensively studied and optimized in biomedicine, and similar technology is poised to be adopted for opera. A first step to realizing this is to amass data in different opera organizations and ultimately include symphonies, ballet companies, museums and other arts organizations. The Statistical Tracking for Audience Growth & Engagement (STAGE) project is the cross-institutional audience analytics platform Opera Verace Foundation is developing to begin meeting these requirements. Initially focusing on opera, STAGE can ingest data about ticket sales, donations, and other data from systems opera organizations use, such as Tessitura or other systems. It then maps the data fields to a common data model we developed (the Performing Arts Ontology for Multi-Organizational Programs, POMOP), and can run procedures and even predictive models for many use cases such as data cleaning, analytics, identifying at-risk high-value patrons, lapsing donors or subscribers, surfacing unexpected attendance patterns, and benchmarking a house’s performance against that of regional, national, or international opera organizations. There is immediate value even if restricted to a single opera house, such as making day-to-day data-intensive tasks more efficient. For prediction and audience understanding, no single opera house has enough data to build reliable predictive models. But many opera houses do have sufficient data, and STAGE makes possible learning from collective experience while keeping each house’s data private through data encryption. We are currently in active partnership discussions with opera houses. #Opera #AudienceAnalytics #DataScience