The Straits Times, Singapore music-makers like Yung Raja and Jack Neo in tune with AI, but wary of its pitfalls

The use of AI in creating music reached a turning point earlier in July when indie rock band The Velvet Sundown – whose members are pure AI creations – went viral and racked up more than one million monthly listeners on streaming giant Spotify.

According to an AFP report, no major streaming service, except for Deezer, clearly labels tracks that come entirely from AI. The producers of these songs tend to be unreachable.

The Velvet Sundown’s Spotify profile states the group are an “ongoing artistic provocation”, a “synthetic music project guided by human creative direction, and composed, and voiced, and visualized with the support of artificial intelligence”.

Closer to home, the use of AI presents both opportunities and challenges for local musicians, songwriters and composers.

Released in April, Raja’s music video was an “experimental” collaboration with Malaysian AI creative studio, Vision Machina.

The rapper had previously used gen AI to come up with album art, but did not end up using it because it did not look good enough.

But when Vision Machina presented a demo video of him rapping – made with gen AI – Raja was struck by how lifelike it was. He agreed to let the company make a full music video for FORREAL, which it fully funded.

“I’ve always stood for innovation, for pushing the creative envelope in ways that people are not used to or have not seen,” the 29-year-old tells The Straits Times.

Using AI helps artistes like him create music videos without being hampered by a lack of budget.

The video has garnered positive comments from his fans, clocking over 250,000 views on his various social media platforms. He has plans to release more AI-generated projects, but is not ready to divulge details.

While excited about AI’s possibilities, Raja is also cautious. He sees AI as a double-edged sword. It offers innovative creative opportunities, but he also understands that it threatens the music industry if it takes away jobs traditionally held by creatives.

His approach is to stay ahead of the technology, constantly learning and finding ways to use AI to elevate his creative processes. “Artistes must adapt to AI or risk being left behind,” he says.

AI-made music also came to the fore at the Singapore International Festival of Arts (Sifa) in May. In Umbilical – a multi-sensorial performance by artists Zul Mahmod, Rizman Putra and thesupersystem – the creators unintentionally ended up with a song with melodies and lyrics resembling a National Day Parade tune while using AI to explore Singapore’s future, a theme of the show.

Zul, 50, says: “During our sessions, we thought it made sense to include AI in Umbilical, since it reflects the times we’re in. So, we dedicated a whole scene to it.”

The song, together with a surreal video produced by thesupersystem, ended up being one of the more memorable parts of the performance.

The sound artist, who has represented Singapore at global events like the Venice Biennale in 2007, has not used AI in his other works, but sees it as a creative tool that can offer fresh musical directions, acting almost like a collaborator.

Having said that, he adds: “If I start relying too heavily on AI, there’s a risk of drifting away from my personal voice and instincts as an artiste. The danger is that the work becomes more about the tool than the intention behind it.

“That’s why, for me, AI is best seen as an extension of the creative process, a way to expand the possibilities, but not something that replaces the human element. It should support the artistic vision, not override it.”

AI also featured in music used on the big screen.

For almost all the songs in his most recent movie I Want To Be Boss, which opened in Singapore cinemas in January, film-maker Jack Neo wrote the lyrics but used AI to come up with the Chinese-language comedy-drama’s music.

The 65-year-old says: “Through these experiences using AI, I’ve been able to clearly see the differences between what AI can offer versus what a human can create.

“Ideally, I had hoped that humans could fully control AI systems to create the perfect song. But in reality, it’s not that straightforward.”

Neo will use the technology to compose a new birthday song in collaboration with local karaoke chain HaveFun, which will be part of the song library at its outlets.

But, like with the songs in I Want To Be Boss, the result will most likely be a combination of AI and human creativity.

“With human composition, there’s something deeply personal which we call the ‘human spirit’, that AI simply can’t replicate. On the other hand, AI has its own kind of ‘spirit’ or character. It really depends on what kind of results you’re looking for, and where or how you plan to use the final piece,” says Neo.

In recent years, academics at the National University of Singapore (NUS) and Singapore University of Technology and Design (SUTD) have come up with their own generative AI music models.

Associate Professor Wang Ye from NUS, for example, used AI and music to aid stroke rehabilitation through a system called CocoLyricist in 2024. CocoLyricist helps patients write and sing their own songs, and supports patient recovery by strengthening brain connections through songwriting.

At SUTD, researchers created Mustango, a text-to-music AI model capable of interpreting detailed musical instructions such as genre, tempo and chord progression, and generating a wide range of music styles. Launched in 2023, Mustango was one of the first models to offer true musical control through text prompts.

Associate Professor Dorien Herremans from the Information Systems Technology and Design pillar at SUTD – where she is also director of its game development team Game Lab – says Mustango has been well received in the music tech community. It has been cited by close to 100 research papers, and its code has been downloaded thousands of times every week.

She says: “First, anybody who needs a musical fragment can use Mustango freely, be it for a video, as an audio sample in a song or just for fun. Second, any researchers or AI developers building generative music models can learn from our published paper and use our code as a basis.”

Mustango is intended as a co-creative tool, she stresses, and not a replacement for musicians. It was trained only on Creative Commons data and its code is open source.

“There is an urgent need for the music industry to implement new mechanisms that stimulate technological progress and allow the use of available training data, perhaps through new attribution methods. We are currently developing such plagiarism detection and similar fingerprinting metrics in the lab.”

BandLab, a Singapore-based online music-making platform with more than 100 million registered users globally since its establishment in 2015, features a free AI-powered idea generator, SongStarter.

Unlike other online music-making platforms such as Suno and Udio, BandLab does not use AI to generate music independently. Instead, its tools are built to support and enhance human creativity, not replace it.

BandLab founder and chief executive Kuok Meng Ru says: “Our artistes use it in different ways. Some use it to break out of writer’s block, and others use it to explore sounds they wouldn’t usually try.

“The most interesting thing we’ve observed is that users who start a song with SongStarter are 80 per cent more likely to finish it. That suggests the tool is doing exactly what it was meant to do – not destroying the creative process, but making it easier to get started.”

BandLab also offers AI tools that do tasks such as removing background noise and turning sounds into usable music data.

“These tools are narrow by design and there to help creators keep moving when the process might otherwise stall,” says Mr Kuok.

“None of them are required. You can use them or ignore them. That flexibility is very intentional.”

BandLab supports the Human Artistry Campaign, a global organisation that advocates for the responsible use of AI in creative industries.

Says Mr Kuok, who is the group chief executive and founder of Caldecott Music Group, which includes companies such as music instrument chain Swee Lee and media outlets like NME: “We deeply believe that human creativity must remain at the centre of music-making, and that core value shapes the way we approach and navigate the world of AI.

“It’s about making sure that as technology moves forward, the rights and role of the creator aren’t left behind. We’re focused on supporting great artistes, and great artistes should always be human.”

While AI-generated music is growing globally, the current impact on home-grown music-makers is limited due to the small size of Singapore’s music industry, says Mr Edmund Lam, chief executive and director of the Composers and Authors Society of Singapore (Compass).

“So far, the impact of generative AI on Singapore’s music industry has been modest. Local creators have not experienced widespread job displacement or high-profile misuse of their works. This may reflect the small scale of our market.”

However, this situation is unlikely to hold, he cautions. Streaming platforms now offer AI-curated playlists, while brands are using AI for advertising and social content.

“These developments threaten to displace human composers in sectors such as background music, gaming, advertising and content scoring – areas where many local musicians earn essential income.”

To address fairness, Mr Lam says Compass is working on ensuring transparency from AI developers, and is developing guidelines on how AI-generated and AI-assisted works should be treated, especially in terms of copyright, attribution and royalty distribution.

“To promote transparency and consumer awareness, original copyrighted works used in AI training should be labelled and disclosed. Creators whose works are used should be credited, and databases of training materials should be made publicly accessible.”