In this piece, you’ll learn:
  • Why lumping every AI-orchestral project under “AI-generated symphonies” misses what makes each one different.
  • How the Beethoven X team used sketches, musicology, and AI to complete the 10th Symphony
  • How later projects — Silicon, FLOW Symphony, already and not yet — use AI in fundamentally different ways
  • The three questions that matter when evaluating any AI-assisted musical work
  • Why audience perception and artistic judgment still decide whether these pieces succeed
  • What orchestras owe audiences in terms of transparency about how AI shaped a piece
When the Beethoven Orchestra Bonn performed Beethoven X in October 2021, listeners faced a question beyond the usual questions like whether they liked the music. They were hearing unfinished ideas from Beethoven, completed by musicians, scholars, and AI. This created a whole new set of problems: how much of the music was really Beethoven, and how much was shaped by today’s people and technology?
Five years later, this question is even bigger. AI now shapes orchestral music in many ways, from new pieces and AI-generated scores to systems that work with live performers. Beethoven X is still important, but it is just one example. New projects aim to expand a composer’s style, create new music, or change how the orchestra sounds. Calling all of these ”AI-generated symphonies” misses what makes each project unique.
A split illustration: on one side, Beethoven’s handwritten sketch fragments; on the other, a modern orchestra performing on stage — visualizing the gap Beethoven X tried to bridge.

What Beethoven Left Behind

Finishing Beethoven’s Tenth Symphony from his unfinished pieces was especially challenging. Beethoven left behind only quick notes, and large pieces were missing altogether. These small bits gave people a few clues about the music, but they didn’t say where each part should go, how the music should change, or how it should sound. So, people had to use their own ideas and imagination to put the music together.
This is not just an AI issue. In 1988, musicologist Barry Cooper created a version of the first movement using the sketches and his understanding of Beethoven. His work shows the main issue: research can limit the possible choices, but it cannot uncover decisions the composer never wrote down.

How Musicians and AI Built Beethoven X

The Beethoven X team used AI and their musical knowledge to complete the Tenth Symphony. The team was made up of an AI expert, a composer, a music teacher, a musicologist, and the project leader. They started by reviewing Beethoven’s notes and trying to figure out how the musical pieces could fit together. After that, the computer filled in the missing parts, created new tunes, and organized everything into one complete piece.
The project took a long time, and the team faced many problems. At first, the music sounded too modern, so the sponsor, Deutsche Telekom, asked the team to make changes. They kept trying different things until the music sounded right. The final version was a group effort with many steps and decisions, not a one-and-done perfect answer. Something similar happened in 2019, when Huawei worked with composer Lucas Cantor to finish Schubert’s Eighth Symphony. The AI created some of the music, and Cantor turned it into a full orchestral score.

Silicon Turns an Orchestra’s Recorded History Into New Music

Since Beethoven X, new projects have gone further than just recreating music from the past. Robert Laidlow’s Silicon, performed by the BBC Philharmonic under Vimbayi Kaziboni in October 2022, used AI in a major modern orchestral piece.
In the third part of the piece, the orchestra played music while also using computer-created sounds made by PRiSM SampleRNN. The system listened to over 2,500 hours of the orchestra’s past music, so the new piece included sounds from its history. Instead of trying to copy what a past composer might have done, the composer used these computer sounds as part of his own music.
What mattered most was how these sounds fit into the piece. For example, if they made the music more exciting, made familiar instruments seem new, or added something different to the whole work. In this case, AI was just a tool for building the music, not a means of evaluating the quality of the work.

FLOW Symphony Uses AI to Respond to Live Musicians

An orchestra mid-performance with a subtle overlay of flowing waveform or neural network lines woven through the string section — representing AI “listening” and reacting to live musicians, as in FLOW Symphony.

Tod Machover’s FLOW Symphony tried something different. The Sejong Soloists first performed it in Seoul in 2024. The piece used string instruments, electronic sounds, and a special computer program trained on sounds Machover made himself. Inspired by rivers, the music mixed real sounds from nature with the instruments. The interesting thing is that the AI could react to the musicians as they played, and listeners online could even change parts of the music themselves.
AI can help make music feel special for different people, concerts, or ways of listening. This is different from writing one set version of a piece. It shows that AI in music can help change things up and make every performance unique, not just help write the music. The people making the music have to decide what the computer can change, when it should help, and what parts should always stay the same.

A conductor’s podium with a musical score on the stand, where part of the page shows traditional notation and part shows a glowing digital/generative pattern — representing the blend of human and AI-composed passages.

AI Compositions Reach Carnegie Hall

Some projects focus even more on music created by AI. At the October 2024 China Now Music Festival at Carnegie Hall, the program included an AI Suite from the Central Conservatory of Music’s AI composition system. The suite had two pieces, “Welcome” and “A Thousand Miles of Rivers and Mountains,” both from 2022. The festival said the suite was partly written by AI, and the notes credited the individual works to the system.

BSO to Field a live AI System

AI compositions continue to appear. In October 2026, the Boston Symphony Orchestra will premiere Machover’s already and not yet, conducted by Kent Nagano. This three-movement symphony uses a custom live AI system to expand and change the orchestra’s sound. It is another case of a well-known composer using AI in his music. Since the performances have not happened yet, we do not know how audiences will respond.
A concert program booklet open to liner notes, with a few lines highlighted — representing the piece’s call for transparency about what came from AI versus human hands.

How Should We Judge AI in Orchestral Music?

All these projects show that we can’t frame AI-assisted composition as humans against AI. Each one uses different ideas, tools, and ways of working. An AI making up a tune, creating an electronic sound, or helping redo an old piece of music are very different things, so we shouldn’t judge them all the same way. Also, just because AI is involved, it doesn’t mean computers can now write symphonies as good as the best human composers. People still need to hear the music to decide if it’s good, not just a frivolous use of technology.
Every time an AI work appears, three questions come up. Does the music sound like the composer? Does it work as a piece of music? Is there proof that the composer wanted these choices? Listeners might agree with the first two but not be able to answer the third. Just because the music is moving does not mean the original composer would have made the same decisions.
At the time, audiences listening to short Beethoven X excerpts couldn’t tell where the original phrases ended and the AI-generated ones began. This is interesting, but it doesn’t prove the whole work is artistically strong. A smooth transition says little about whether a long movement develops its ideas well or ends satisfyingly. Here, people’s opinions and feelings really matter. Someone has to pick which musical ideas are interesting, which parts get boring because they repeat too much, and when the music should end. These choices aren’t about ignoring what computers can do—they help make each performance special in its own way.
Orchestras can help people in the audience by explaining how the music was put together. The concert program should note which parts came from old music, which were made by computers, and which were created by the musicians playing now. New pieces should also explain exactly how AI helped, rather than just mentioning AI in general. This way, everyone can understand and decide what they think about the music.

Conculsion

Whether these music projects turn out well depends on how artists use their tools. A piece of Beethoven’s unfinished music can help someone create something new and powerful, without pretending that it is exactly what Beethoven would have written. An AI system can help the orchestra do more, but it won’t replace a real composer. AI’s growing presence in orchestral music asks us to be more precise about what we hear and whom we credit.

Resources: