What if your notation software could sing? At the top of this episode, we play a short clip, performed entirely in Dorico with NotePerformer handling the orchestra, and a plugin called Cantai rendering the baritone voice.

That voice is synthesized directly from the Dorico score with minimal configuration, and it marks the arrival of something the notation world has been waiting for for a long time. Philip Rothman and David MacDonald talk with Richard deCosta, composer, software developer, and founder of Cantai and the Turing Opera Workshop, about what it took to build it.

The conversation goes deep on the technology: why synthesizing the voice is fundamentally harder than synthesizing instruments, how the phonemizer works, why Cantai renders offline rather than in real time, and what it really means to build a plugin that reads a score rather than simply receiving MIDI.

Richard also explains how years of frustration with the disconnect between notation and external vocal synthesis tools — from EWQL Symphonic Choirs and WordBuilder to ACE Studio and Synthesizer V — led to the central insight behind Cantai: that the lyrics were always there in the score; they just weren’t being passed to the playback engine.

We also dig into the ethical and business model Richard has built around the singers whose voices power Cantai. Every vocalist is contracted, paid a competitive recording fee, and receives an ongoing share of the product’s profits in proportion to how much their voice is used. Cantai is already live for MuseScore Studio and Dorico, and arriving for Sibelius on May 30. The roadmap — more languages, less vibrato, Broadway and jazz styles, and a thought-provoking vision for the future of real-time vocal generation — gives us plenty to look forward to.

Directly mentioned or closely related: