Last year, amid the explosion of the AI music market, we wrote about $4.8 billion-valued technology giant Kunlun Tech and its launch of a fully AI music streaming service and a music generator called Mureka.
Now Kunlun Tech, whose market cap has grown to around 43 billion Chinese yuan ($6 billion), has unveiled what it’s calling the “world’s first music reasoning large model.”
Kunlun claims that Mureka O1 outperforms competing models like Suno V4 across multiple metrics, with strong results in mixing quality, vocal textures, and background instrumentation. The company says its AI models’ ability to carry out subjective assessments places them “among the top-tier” in the category.
The new Mureka V6 and O1 models support a range of musical styles like jazz, electronic, pop, country, R&B, rock, and dance. Both models also offer emotional expressions including “happiness, indulgence, mystery, vitality, sadness, and various other emotional expressions,” according to a press release.
Notably, Mureka introduces two features that could change how artists and producers interact with AI tools. The first is a “Music Reference Function” that allows users to upload existing audio files or YouTube links as creative references.
The second is Mureka’s vocal function. The company claims to be “the world’s first AI music generation platform that allows users to choose the vocal timbre of the singer.” In addition to selecting from pre-programmed vocal styles, users can upload their own voices for the AI to learn and copy, creating what Kunlun describes as “personalized exclusive works with a single click.”
The platform includes a Mureka Store where users can “sell AI-generated music”.
Both AI models have been made available through open API services, becoming “the world’s first high-quality AI music generation platform” to do so, Kunlun’s AI unit Skywork said. This way, developers and music platforms will be able to integrate Mureka’s capabilities into their own products and services.
Kunlun is the parent company of web browser Opera and the former owner of Grindr. The company’s Skywork unit was developed by Kunlun Tech in partnership with SINGULARITY AI. At the time of its launch, Skywork was described as being “benchmarked against ChatGPT” and capable of “interact[ing] with users through natural language.”
Most recently, the Skywork division announced that Mureka O1 introduces “chain-of-thought” (CoT) technology to music generation for the first time. This technology incorporates thinking and self-criticism in the inference process, the company said.
“Unlike traditional autoregressive models that generate audio step-by-step, MusiCoT pre-generates the overall music structure before finalizing the audio tokens for decoding. This approach significantly improves the structural coherence and instrumental arrangement of the generated music,” said Skywork.
“MusiCoT pre-generates the overall music structure before finalizing the audio tokens for decoding. This approach significantly improves the structural coherence and instrumental arrangement of the generated music.”
SkyWork AI, Kunlun Tech Unit
The company further explained that MuisCoT no longer requires manual annotations. “It not only enhances the analyzability and quality of generated music, but also paves a new path for high-fidelity music generation.”
Additionally, Kunlun is offering model fine-tuning based on Mureka V6, offering customized AI music services for musicians, producers, brands, and game developers.
In addition to its AI music models, Kunlun also launched a fully AI music streaming service in August last year. The service called Melodio features “personalized, AI-generated music streams tailored to [users’] moods and scenarios.”
Most recently, Kunlun Tech Chairman and CEO Fang Han hinted at the company’s plan to continue increasing investment in AI models, particularly music generation models globally, China’s Yicai Global reported on Thursday (March 27).
Skywork noted that the Mureka platform, a wordplay on “Music” and “Eureka,” has attracted users from more than 100 countries since its initial launch. The V6 model supports music creation in 10 languages, including English, Japanese, Korean, French, Spanish, Portuguese, German, Italian, Russian, and Chinese.
Users worldwide can now access the new V6 and O1 models through Mureka’s website.
Kunlun’s push into AI music technology comes amid increasing competition. Other tech giants have also entered the space including ByteDance, which launched mobile audio creation tool Sponge Band, while Tencent Music Entertainment’s Tianqin Lab has collaborated with Tencent AI Lab on its own music AI model.
It also comes at a critical time for the AI music space. Competitors like Suno have faced significant legal challenges, with major record labels suing the company for using copyrighted recordings without permission.
In June, the $500 million company was sued by the major record companies, along with fellow AI firm Udio, for allegedly training their systems using the majors’ recordings without permission – an accusation they pretty much admitted to in court filings in August.Music Business Worldwide
