Puccini · T4T · 2026
The Four Turandot
Four theatres on three continents, four endings, and an opera that rises again in virtual reality. Here the Miller Theater in Augusta builds itself around Calaf.
The Four Turandot · T4T · 2026
Four endings
Puccini died before finishing Turandot. The opera has four endings, and T4T staged one in each city. Pull the thread of Act Three: at Liù’s death you can stop, as Toscanini did in 1926, or choose an ending.
Pull the thread, or choose a city.
Drag the thread with your finger or the mouse. From the keyboard: arrow keys and Enter, or the four buttons below.
In the background, our render of the Miller Theater in Augusta.
The project in brief
Puccini beyond the centenary
The Four Turandot is an advanced artistic training project born around the centenary of Turandot: the world première was conducted by Arturo Toscanini at La Scala on 25 April 1926. Between February and April 2026 the opera was staged in four theatres, in Italy, the United States, South Korea and China, each time with a different ending. The closing concert, on 8 May in Milan, set the three endings side by side.
The lead institution is the “Giuseppe Verdi” Conservatory of Milan. The partners are the Conservatories of Florence, Modena and Carpi, Piacenza and Turin, the Politecnico delle Arti of Bergamo, the University of Milan and NABA, Nuova Accademia di Belle Arti. Casa Ricordi and the Archivio Storico Ricordi have joined the project. Abroad, the performances were hosted by the Miller Theater in Augusta, Keimyung University in Daegu and the Zhejiang Conservatory of Music in Hangzhou.
GR STUDIO works for the “Giuseppe Nicolini” Conservatory of Piacenza, one of the partners: on its behalf, GR STUDIO is the project’s scientific lead.
The official project website: t4t.consmi.it
Project funded by the European Union, NextGenerationEU, within Italy’s recovery plan (PNRR).
The four stages
Four theatres, four endings
The cast was double, so each city had two performances: in Augusta on the same day, 7 February. Each city had its own ending. Dates and venues link to the news items on the official website.
Augusta Georgia, United States
Miller Theater
The version interrupted at Liù’s death, as at the 1926 première.
The official news itemMilan Italy
Sala Verdi, Conservatorio di Milano
Luciano Berio’s ending (2001).
The official news itemDaegu South Korea
Keimyung Art Center
Franco Alfano’s second ending.
The official news itemHangzhou China
Zhejiang Conservatory of Music
Franco Alfano’s first ending.
The official news itemClosing concert “Puccini oltre il centenario”
Milan, Sala Verdi, Conservatorio di Milano
Alfano’s two endings and Berio’s, one after the other, on the same evening.
The official news item
The four endings, in brief
Puccini died on 29 November 1924. He had written and orchestrated the opera up to Liù’s death and left 36 pages of sketches for the rest. Franco Alfano completed the ending from those sketches: a first, longer version, then a second, shorter one, which is the one that entered the repertoire. At the première, on 25 April 1926, Toscanini stopped the performance at Liù’s death; from the second performance Alfano’s ending was played. In 2001 Luciano Berio wrote a new ending from the same sketches.
Enter the Miller
Scroll, and you walk into the theatre.
- You arrive from the street, under the marquee: “Puccini Turandot”.
- The foyer, the box office, the Turandot posters.
- The doors with porthole windows, the red carpet.
- The stalls, and at the far end the stage.
Our work
What we brought home
From each performance we collected four things, to reconstruct the opera in a virtual environment that can also be walked through with a headset.
Sound in space
Ambisonics recordings, which keep the position of the orchestra and of the performers as they move.
The acoustic signature of the theatre
The impulse responses of each hall, to make the virtual theatre sound like the real one.
Gesture
The performers’ movements, captured with an infrared OptiTrack system and with 360° cameras.
Faces and costumes
Scans of the performers of both casts, in make-up and stage costume.
A tour begins with the technical rider: what the host theatre needs, what we bring, how two technical crews that do not share a language talk to each other.
The eight duties of the assignment
The “Giuseppe Nicolini” Conservatory of Piacenza entrusted GR STUDIO with the scientific responsibility for the project. The assignment is divided into eight duties.
- Project direction
- Software programming
- 3D scanning of the spaces
- Modelling of halls and faces
- Acoustic sampling of the spaces
- Coordination of the live events
- Production coordination
- Research and tests with the partners
How it is made
From the performance to the scene
The chain that brings an evening of theatre inside an environment you can walk through: five steps, from the real hall to sound that moves with the singer.
In the theatre, before the show
Photographs and infrared scans of the architecture. The performers’ faces scanned with their stage make-up on, so that the digital character looks like the one the audience saw that evening. In the hall, an OptiTrack PrimeX 41 optical system for markerless motion capture — no suit, no markers.
The sound of the space
Ambisonic recording, which preserves the position of the sources, and measurement of ambisonic room impulse responses throughout the hall, from several positions on stage: the theatre’s acoustic signature.
From video to movement
The infrared footage is synchronised and calibrated in OptiTrack Motive; then Captury reconstructs the body motion from the multi-camera video, without markers on the performers.
Faces, costumes, halls
In Blender the scans become clean, light heads: retopology, UVs, textures, normal maps. The same work goes into the theatres, from the surveyed geometry to surfaces derived from photographs — always with an eye on weight: everything has to run in real time.
In Unreal Engine 5
Face and motion enter the MetaHuman pipeline; the lip sync comes from the vocal track and the costume, designed by NABA Milano, is simulated as fabric. The voice moves with the singer, the orchestra stays in its place on stage, and over everything the measured impulse responses are applied.
The theatres in 3D
The halls stand up again
After China the work moved to the laboratories of XNL Musica and of the Conservatory, in Piacenza. The theatres are reconstructed in three dimensions, stalls, gallery, proscenium and fly grid, and the performers return to the stage as digital characters. The first was the Miller Theater in Augusta.



The leading roles, in close-up
Turandot, Calaf, Liù and the emperor Altoum, rebuilt in Unreal Engine 5. Faces start from scanning the performers in full stage make-up; then they are rebuilt and dressed in materials, skin, hair and metal, and the costumes are simulated as real fabric. Now they are finished: characters that can hold a close-up.




Two halls, two routes
The first two halls were born in different ways: Augusta from the theatre’s CAD model, Hangzhou from scratch, from floor plans and photographs. The show lighting is a test: the real lighting belongs to whoever makes the show.
Augusta, rebuilt on the theatre’s model
The Miller Theater was dressed again from the ground up: seats redrawn on the official plan, carpet and floors, acoustic panels, a truss with moving heads, recessed spotlights, the gallery handrail, L-Acoustics KARA speakers flown at the sides of the proscenium.




Hangzhou, built from scratch
The hall of the Zhejiang Conservatory of Music had no model: it was built from DXF floor plans and from photographs taken in the theatre. Measured on site: the pit rail, the back half of the stalls, the seats per row; estimated: the gallery front, the upper boxes, the ceiling and the colours.




The Miller Theater, from wireframe to hall
A minute and a quarter retracing the stages: first the wireframe of the model, then the volumes of the stalls and the balcony, the seats row by row, the materials, the lights and the murals beside the proscenium. It is the order in which every theatre comes to life: geometry first, then matter, light at the end.
Entering the Miller Theater
You come in from the street, under the Miller’s marquee with “Puccini Turandot” in lights: the box office, the corridor with the Turandot posters, the doors with porthole windows, the red carpet. From here there are two ways: one leads to the stalls, the other climbs the stairs to the balcony. It is the same theatre rebuilt for virtual reality, seen through the eyes of someone arriving.
Act Two: Ping, Pang and Pong
Act Two opens with the Emperor’s three ministers, Ping, Pang and Pong. They don’t yet know whether the stranger will win the trial of the riddles, so they prepare for both: the wedding and the funeral. Then they sit down, tired of a China where heads keep falling, and dream of their homes far away. These are the first four scenes of the tableau: the orchestra on stage behind them, a single follow spot, and the blue sashes that tie them to one another. Every scene was rendered from the stalls and from the gallery; here we show one moment at a time.

Scene 1The entrance, from the gallery: the blue sashes run across the stage from one to the next. 
Scene 1Red lanterns for the feast, white ones for mourning. 
Scene 1The paper strip, with the lanterns set down on the floor. 
Scene 2The scroll, read by all three. 
Scene 3Sitting on the floor, the sheet in his lap. 
Scene 3The open sheet, from the gallery: the light holds them close. 
Scene 4On their feet again, with the hanging sheet. 
Scene 4Around the little table, with a fourth, blindfolded figure.
One staging, four theatres
The four halls are now complete: around Calaf there are the orchestra and the chorus, and behind him the projected backdrops. The staging is the same in all four theatres, and the backdrop is the sky of the night in which no one sleeps: the stars that Calaf, at the end of the aria, asks to set. What changes is the theatre around it, and with it the way the stage is seen. Each hall is seen from the back of the stalls and from the side, from the seats of those who walk through it with a headset.
Augusta Miller Theater
The wide proscenium holds everything: the orchestra and the chorus in silhouette on stage, behind Calaf, and the sky running along the walls to the sides of the stage, next to the murals.

From the back of the stalls. 
From the side, towards the proscenium mural. Milan Sala Verdi of the Conservatory
It is a concert hall, with no proscenium: the backdrop is not behind the stage but on the walls of the hall, and the chorus sits on the tiers at the back. Here the whole hall becomes the stage.

From the back of the stalls: the walls become the backdrop. 
From the side, from the front rows. Daegu Keimyung Art Center
A theatre with a proscenium: the backdrop closes the stage on three sides, like a box of light, and the chorus remains a row of silhouettes at the back. Calaf is further away, and the stage is seen like a painting.

From the stalls: the backdrop closes the stage on three sides. 
From the side, towards the stage. Hangzhou Zhejiang Conservatory of Music
The same box, in a deeper hall with a balcony. From up there the stage reads as a whole: the backdrop on three sides, the chorus in silhouette and Calaf in the rectangle of blue light, above the orchestra pit.

From the balcony: the whole stage. 
From the back of the stalls. 
From the side, from the front rows.
“Nessun dorma”
The whole aria, three minutes and thirteen seconds, while the theatre changes around Calaf: it starts in the Miller Theater in Augusta, moves through the Sala Verdi of the Milan Conservatory and the Keimyung Art Center in Daegu, and ends in the hall in Hangzhou. The staging is complete, with the orchestra, the chorus and the projected backdrops, and the camera also goes up on stage, behind Calaf, to look at the hall as he sees it. Calaf, the stages and the halls are entirely reconstructed in Unreal Engine 5. Calaf is a MetaHuman animated with the motion capture data recorded during the performance; the lip sync comes from the vocal track and the costume is simulated as fabric, in real time.
Crossing sound
The hall’s real acoustics, on headphones
Alongside the virtual reality platform we are writing a second piece of software: an audio renderer that takes the impulse responses measured in the theatres and applies them to the virtual environments. The voice is produced by convolution with the hall’s ambisonic impulse responses, the ARIRs, which describe how it reflects off the walls, the stage and the stalls. On headphones it becomes binaural, and it changes with the listener’s position and orientation.
The graphics engine sends it the listener’s position up to 60 times per second. The renderer drives the scene time, and on that time animation, lip sync and voice stay together.
On 25 September 2026, at the European Researchers’ Night in the former Carmine church in Piacenza, we let people try it: with a joystick and a pair of headphones you could walk through the stalls of the Miller Theater while Calaf sings “Nessun dorma”, with the hall’s real acoustics. One person at a time, from 8:30 to 14:00, aged fifteen and over.
At the Carmine, 25 September 2026



An experiment: finding the voice in the dark
On the same setup we built a localisation test. You start from a random point in the stalls, facing the stage; then the image goes out completely. No references: neither the hall nor the singer. You have to reach the point that sounds closest to the voice, guided by hearing alone.
We record whether you get there, how long it takes, the path travelled and the final position: these are measures of how well a source is localised, and of the idea of the hall one builds from listening alone. The data are being analysed.


Where it comes from
From Georgia to Piacenza
The method predates T4T. In February 2023 we began a research partnership with the University of Georgia, in Athens: to rebuild the sound and the rooms of the Dancz Center in virtual reality — not only how they looked, but how they sounded. More than nine thousand photographs in two months, rooms explorable with a headset in Blender and Unreal Engine 5, the first tests of binaural and ambisonic sound.
From October 2023 the work moved on to people: face scanning and motion capture of the body, facial expressions and hands, to bring a performing artist into virtual reality together with their music. In November 2024 the Milan and Piacenza delegation visited Athens, and the University of Georgia became a research partner on the software: within the partnership our lead researcher holds the title of International Collaborator at the university. As Libertà reported on 14 April 2026, the heart of the project is “an experimental software developed after three years of research with the University of Georgia”. In May 2026 the leadership of the Dancz Center came to Piacenza for days of joint review, on method and on code.
Credits and links
To find out more
This is not the project’s official website. It is a GR STUDIO SRL website about the work carried out for the “Giuseppe Nicolini” Conservatory of Piacenza, a partner of The Four Turandot. The photographs of the performances are on the official website: here we link to the news items, without copying them.
The project
Our account
Funding
THE FOUR TURANDOT – T4T, INTAFAM_00048, CUP F41I23000550001, PNRR Missione 4 Componente 1 Investimento 3.4, NextGenerationEU



