■ SAFE RISK ■ Sports & Entertainment
No. Synthesis can imitate the overtone spectrum of khoomei convincingly — that is a solved audio problem — but throat singing is a living cultural practice performed by people from specific traditions, and audiences and festivals book the tradition, not the timbre.
“Producing two notes simultaneously with your throat. AI can synthesize it. But perform it? No.”
Our AI replacement risk score — how we score jobs
The craft is physiological and slowly acquired. Techniques such as khoomei, sygyt and kargyraa involve controlling the false vocal folds, tongue position and pharyngeal shape to split a fundamental into audible overtones — a skill built over years and traditionally learned by ear from an elder. Working throat singers in Tuva, Mongolia and beyond perform at festivals and concerts, tour internationally with ensembles, teach workshops, record for film and games, and often play igil or morin khuur alongside singing. Many hold the practice as an inherited cultural role rather than a job.
Audio technology can already imitate this. Spectral synthesis and neural voice models reproduce overtone singing well enough for a soundtrack bed, and library sample packs of Mongolian and Tuvan vocals have existed for years. That competition is real for one specific revenue line: incidental music for media, where a producer who needs eight seconds of ethnic-sounding vocal will reach for a sample. Meanwhile AI helps elsewhere — mixing, distribution, translation for international touring, and archival preservation of elder singers' recordings.
What is not substitutable is the performance economy this actually runs on. International touring by Tuvan and Mongolian ensembles is built on cultural authenticity — audiences and festival programmers are buying access to a tradition, and the artists' identity is inseparable from the product. UNESCO-recognised intangible heritage status channels funding toward transmission by living practitioners. Teaching is entirely embodied; you learn khoomei by having someone listen to your throat and tell you what to change. Ceremonial and community functions are not performances for sale. And the visceral experience of hearing two pitches emerge from one person in a room is a physical event. Our score of 10 reflects a practice where the sound is copyable and the practitioner is not.
Automatability: our editorial assessment of current and near-term AI capability
Sample libraries eroded the incidental-media market years ago and neural synthesis finishes that job now — a small slice of income, already largely gone. Live performance, touring and teaching stay intact well past 2040, supported by heritage funding and international festival demand. The genuine long-term risk is transmission: whether enough young singers learn the technique. That is a cultural continuity question, not an automation one.
The sound, yes. Overtone spectra are well understood, sample libraries of Tuvan and Mongolian vocals have existed for years, and neural voice models now generate convincing imitations. That has already taken most of the incidental-soundtrack work. What synthesis cannot supply is a performer from the tradition standing on a festival stage, which is what the touring economy actually sells.
For those who commit to it seriously, yes — international touring, festival circuits, teaching and heritage funding sustain a working ensemble scene, particularly for Tuvan and Mongolian groups. The income is concentrated in live performance and instruction. Recording royalties and media placement are the weakest and most automation-exposed lines, and were already thin.
Less directly than people assume. Synthesis competes for background-music budgets, and generative models trained on traditional recordings raise real questions about consent and benefit-sharing. But live cultural performance, ceremony and person-to-person transmission are unaffected. The greater threat to these traditions has always been demographic — fewer young people learning from elders.
Prioritise live work and touring, since that is where cultural authenticity commands a premium. Develop teaching, including online instruction, which reaches a worldwide audience fascinated by the technique. Seek heritage and transmission funding, which explicitly supports living practitioners. And collaborate outside the tradition — cross-genre projects have introduced khoomei to audiences a sample pack never reaches.