AI Can Already Write Songs. How Will Musicians Make a Living? What Hatsune Miku Tells Us About the Future of Japanese Musicians

Key Points

• AI can already write melodies, lyrics and vocals all at once. But Japan’s music world went through a similar shock in 2007, when the singing synthesizer software “Hatsune Miku” meant songwriters no longer needed to find a singer. Singers didn’t disappear. The biggest winners were composers who published their work online and went on to fame, such as Kenshi Yonezu and Ayase of YOASOBI.
• In Japan, JASRAC, which collects and distributes music royalties, only manages songs with “human creative contribution.” Songs generated automatically by AI from simple prompts aren’t treated as copyrighted works, so no one can collect royalties on them.
• Japan’s music income is hitting new highs, and the growth is concentrated where “this particular person” matters. In the first half of fiscal 2026, JASRAC distributed 74.14 billion yen to rights holders, and the live entertainment market reached 760.5 billion yen in 2024, driven by fans spending to support specific artists.
• The first to feel price pressure from AI is functional music for videos and ads. What musicians will be able to earn the most from in the future is what AI can’t offer: a name people remember, authorship, and live shows that require the artist to be there in person.

On January 24, 2026, four people stood in a row on the stage of Yokohama BUNTAI. syudou was on drums, Ayase on keyboards, and two others on guitar and bass. They introduced themselves as that night’s opening act, a unit called DREAMERS.

None of the four songs they played had originally been sung by a person. On the giant LED screen behind the stage, 3D images of Hatsune Miku, Kagamine Rin and Kagamine Len took turns appearing and singing the melodies. These characters are the voices of computer software. The four people standing in front of the screen were the composers who had written those songs years ago and had the software sing them.

On the same night, Ayase switched roles and took the stage with singer ikura as YOASOBI. The headliner was syudou, singing himself.

Seen in 2026, the scene is especially striking. Today’s AI can generate a complete song with melody, lyrics and vocals from a single line of text in a few dozen seconds. Many musicians are asking: if machines can even write songs, what’s left for people to sell?

Japan has actually answered a similar question once already. Nineteen years ago, a piece of software made it possible to sing without a real human. Where did the money end up going that time?

In 2007, software meant songwriters no longer needed a singer

Hatsune Miku is “singing synthesizer” software. Users enter a melody and lyrics into a computer, and the software sings them using prerecorded voice samples. The core technology is VOCALOID, announced by Yamaha in 2003. Crypton Future Media, a software company in Sapporo, Hokkaido, turned it into a product and released it on August 31, 2007.

Before that, people who could write songs but couldn’t find a singer had to sing themselves. Singing synthesizers weren’t new either. Crypton’s previous product, MEIKO, released in 2004, sold about 3,000 copies in its first year, in a niche market where selling 1,000 counted as a success.

Hatsune Miku made one crucial change. Crypton’s president, Hiroyuki Ito, says they decided to cast not a singer but a voice actor, someone who voices anime characters. His reasoning was that users didn’t want vocal technique so much as a personality that made them want to make her sing. The voice was ultimately sampled from voice actor Saki Fujita, and the package featured a 16-year-old girl with green twin tails.

The software sold about 30,000 copies in six months. The first to embrace it weren’t professional musicians, but amateur composers, people who liked the character, and younger listeners. Many veteran musicians actually kept their distance at first because of its anime-style look.

One thing should be clear: Hatsune Miku can sing, but it can’t write. Every melody and every line of lyrics was still written by a person. Today’s AI goes a step further and can write songs too. So this history isn’t an exact rehearsal. But it shows very clearly one thing: when one link in the chain suddenly becomes cheap, where the money moves.

Once machines could sing, songwriters were suddenly in demand

After Hatsune Miku went on sale, huge numbers of songs were uploaded to Niconico, a Japanese video site where viewers’ comments scroll across the video in real time. The people who wrote these songs came to be called “Vocaloid producers,” or “Vocaloid-P,” with the P standing for producer.

In 2009, a young man calling himself Hachi began posting songs sung by Hatsune Miku on Niconico. One of them passed 100,000 views within three days of being uploaded. From 2012 he started using his real name, Kenshi Yonezu, and singing his own songs in his own voice. In 2020 his album STRAY SHEEP topped Billboard JAPAN’s year-end album chart.

Ayase’s path came nearly a decade later. He had been the vocalist of a rock band and didn’t post his first Vocaloid song until December 24, 2018. The following year, staff from the novel-posting site monogatary.com approached him with an experimental project: turning novels posted by users into songs. Ayase formed YOASOBI with singer ikura, and they debuted at the end of 2019 with “Yoru ni Kakeru” (“Racing into the Night”), which went on to top Billboard JAPAN’s year-end song chart.

The software solved the “no singer” problem for them and let their work be heard directly. But what made them famous was the recognizable name behind the work. Listeners didn’t remember “that song Hatsune Miku sang.” They remembered “Hachi’s new song.”

The machines didn’t take singers’ jobs. They raised a new generation of singers

If the story ended here, it would sound like “the machine replaced singers and made composers.” But what actually happened is more interesting.

Once Vocaloid songs caught on online, a new kind of creator emerged: “utaite,” or cover singers. They re-recorded Vocaloid songs with human voices and uploaded them as “utattemita” (“I tried singing it”) videos. The machine made the song a hit first, and then humans took it over and sang it.

That’s how Ado started. She had admired these cover singers since sixth grade, and in 2017, in her second year of junior high, she uploaded her first cover video. Early on it had only about two dozen views. She later said she wasn’t good at schoolwork or sports and didn’t feel sure she’d be able to hold down a job once she grew up. Cover singers, who didn’t show their faces and were judged only by their voices, made her feel that “maybe here, I could do something.”

In 2020, a Vocaloid producer wrote Ado’s debut song, “Usseewa.” It became such a hit that it made Japan’s top ten buzzwords of 2021. In 2022, “New Genesis,” sung by Ado, became the first Japanese song to top the Apple Music global chart.

The person who wrote “Usseewa” was syudou.

From office worker to the Budokan: how one composer’s income grew

syudou started writing songs with Vocaloid as a student. After graduating he became an office worker and kept writing after work. A song he released in January 2019 took off, and at the end of that year he quit his job to focus on music.

The path over the next few years reads like a map of a musician’s income. In 2020, as a composer, he wrote for Ado, and every time the song was played or sung at karaoke, he as the writer received royalties. In 2021 he started singing his own songs in his own voice. In August 2022 he held his first show with an audience, in Nakano, Tokyo, and tickets sold out the same day. On May 31, 2025, he played the Nippon Budokan, with tickets at 8,300 yen. In January 2026 he organized a two-day music event at Yokohama BUNTAI, the night this article opened with.

With each step, his income gained another source that “only he can provide”: authorship brought royalties, his taste brought commissions, his voice brought his own releases, and live shows brought ticket sales. Hatsune Miku was his starting point, but not one of those income streams came from the software itself.

This time, AI can write the songs too

The biggest difference between AI and Hatsune Miku is that AI has taken over the “writing” as well.

SOUNDRAW, a Tokyo startup founded in 2020, offers an AI composition service. Users choose the mood, theme and length of a video and get 15 tracks in about ten seconds. Its main customers are video creators for YouTube, TV and games.

The company has one detail that’s easy to miss: the music its AI learns from was all created by composers the company hired itself. Founder Daigo Kusunoki says this means the company won’t have to worry even if copyright rules for AI change in the future. In other words, even an AI composition company starts by paying people who can write songs.

Kusunoki’s view of the market is blunt. He thinks the music market will split into two tiers. Functional music in large volumes, like video background music and short ad jingles, will be dominated by AI. Meanwhile, the story behind a song, its personality and the person themselves will become the new measure of value.

That also means AI’s impact won’t fall evenly on everyone. For jobs writing background music for online videos or a few dozen seconds of music for small ads, clients want something that “just works” and don’t much care who wrote it. That kind of work will face price pressure first.

Songs without an author can’t collect royalties

In Japan, when a song is used, the money flows back to creators mainly through JASRAC, the Japanese Society for Rights of Authors, Composers and Publishers. Karaoke venues, concert promoters, streaming platforms and TV stations pay JASRAC to use music, and JASRAC distributes the money to lyricists, composers and music publishers based on usage records.

JASRAC’s rules on AI songs reveal the core of the royalty system. It only manages songs with human creative contribution. Lyrics and melodies generated automatically by AI from simple prompts aren’t considered copyrighted works, and JASRAC won’t manage them. In other words, a song made entirely by AI could be played a million times, and within this system there would be no “author” to share in the money.

And this person-centered system is collecting more money than ever. In the first half of fiscal 2026, JASRAC collected 75.6 billion yen in usage fees and distributed 74.14 billion yen to rights holders, both first-half records. Of that, online use such as streaming brought in 32.73 billion yen, up 8.7% year on year, and karaoke and large dome- and arena-scale concerts were also growing. In the same period, JASRAC sent creators a survey on the economic impact of generative AI and began testing technology to identify AI-generated songs.

There’s even more money in live shows. In 2024, the market for live music and stage performances in Japan reached 760.5 billion yen, with music accounting for 529.9 billion yen, both all-time highs. When research firms analyzed the growth, they pointed to two spending habits. One is “oshikatsu,” spending steadily to support your favorite artist. The other is “toki shohi,” or “moment spending,” paying for moments that can’t be repeated and can only be experienced in person.

Both kinds of spending share one premise: there has to be a person on the other side.

A song will keep getting cheaper. A person won’t

Back to that night at Yokohama BUNTAI. On stage, DREAMERS said the unit’s name was something a few friends came up with over drinks in 2019, and that “our original goal was a trip to Hokkaido, and somehow we ended up at BUNTAI.” syudou, closing the night, told the crowd: “I’m here to show you clearly that I’m an artist who deserves to headline.”

When Hatsune Miku made singing cheap, singing voices stopped being scarce. What became scarce was the person who wrote the songs, the person who picked them up and sang them, and the person everyone was willing to pay 8,300 yen to see. AI is making songwriting cheap too, and that shift will only go further. “Pretty good songs” will become so plentiful that no one cares who made them, and people willing to pay will concentrate their money even more on people they can recognize.

This isn’t a conclusion that leaves everyone safe. Work that only sells “a song that does the job” really will make it harder and harder to earn a living. But for people who want to live off music, Japan’s experience 19 years ago has already pointed the way: use the machine as a tool to get your work heard, and then make listeners remember your name.

That night, on the big screen, Hatsune Miku sang songs from more than a decade ago. What the audience had bought tickets to see were the people in front of the screen.

---

Notes

• Sources differ on whether “Yoru ni Kakeru” debuted in November or December 2019, so this article says “the end of 2019.”
• Sources variously say syudou started releasing Vocaloid songs in high school or in 2012, so this article says “as a student.”
• The figure of about 30,000 copies of Hatsune Miku sold in six months was published by Crypton for its fifth anniversary in 2012. SOUNDRAW’s generation speed and customer base are from the company’s early public information and may have changed.
• syudou’s remarks in Yokohama and DREAMERS’ comments are paraphrased from a live report.

References

1. Billboard JAPAN (柴那典)｜syudou主催【PENTATONIC】1日目レポート｜2026-01-30 (DREAMERS’ members and roles, four Vocaloid songs with LED visuals, the “Hokkaido trip” and syudou’s headliner remark, YOASOBI’s set)
2. BCN (伊藤博之 interview)｜『初音ミク』はこうして生まれた｜2009-07-01 (casting a voice actor rather than a singer, MEIKO’s first-year sales of about 3,000, 1,000 copies as success)
3. ITmedia NEWS｜初音ミク 5 周年｜2012-08-31 (about 30,000 copies in six months, early adopters and professional musicians’ reactions)
4. nippon.com｜初音ミクとボカロの歴史 (released 2007-08-31, VOCALOID2, voice sampled from Saki Fujita, STRAY SHEEP and “Yoru ni Kakeru” topping year-end charts, composers previously had to find a singer or sing themselves)
5. Newsweek Japan｜ボカロ文化と紅白 (VOCALOID announced in 2003, Vocaloid alumni on Kohaku); page 4 (“Usseewa” in the 2021 buzzword top ten, “New Genesis” topping the Apple Music global chart)
6. Ticket Ryutsu Center｜米津玄師（ハチ）のボカロ曲 (started posting in 2009, 100,000 views in three days, switched to his real name in 2012)
7. KAI-YOU｜Ayase (former band vocalist, first post on 2018-12-24, formed YOASOBI in 2019)
8. Bunshun Online｜YOASOBI 誕生の経緯｜2020-12-17 (the monogatary.com project, forming the duo with ikura)
9. Real Sound｜Ado interview｜2021-10 (first cover posted in 2017 in her second year of junior high, about 24 early views, worries about work and why she started)
10. Billboard JAPAN｜syudou feature (started as a student, office worker years, January 2019 hit and quitting at year’s end, singing himself from 2021, first Nakano show sold out in August 2022, Budokan tickets at 8,300 yen)
11. Natalie｜syudou Nippon Budokan show (Nippon Budokan, 2025-05-31)
12. Forbes JAPAN｜SOUNDRAW’s Daigo Kusunoki (founded 2020, 15 tracks in about ten seconds, customers, training data from in-house composers)
13. Music.AI｜AI splits music industry: Daigo Kusunoki, SOUNDRAW (a two-tier market, AI dominating functional music, human stories and personality as value)
14. JASRAC｜AI を利用した楽曲の管理 FAQ (only manages songs with human creative contribution, doesn’t accept automatically AI-generated works)
15. Real Sound｜JASRAC first-half fiscal 2026 business report｜2026-10 (75.6 billion yen collected, 74.14 billion yen distributed, 32.73 billion yen from streaming, AI survey and detection tests)
16. KAI-YOU｜2024 年ライブエンタメ市場は過去最高の 7605 億円｜2025-06 (market size, 529.9 billion yen for music, oshikatsu and moment spending)
17. Agency for Cultural Affairs｜AI と著作権 (official portal on Japan’s interpretation of AI and copyright)