‘Suno’ as seen by domestic firms… “Eventually, music creation AI will even diversify.”


(Photo = Suno AI)

Suno AI's music creation artificial intelligence (AI) service 'V3' is causing a sensation overseas. Even users with no musical knowledge can create a whole song that features each musical instruments and vocals in only just a few seconds just by entering text, and it has been praised because the 'music version of Sora.'

In response to this, Newtune (CEO Jong-pil Lee, CEO) and Poja Labs (CEO Won-gil Heo), representatives of domestic music creation, respectively tested Suno and agreed that “there are clear differences between music creation technologies.”

Newtune is an organization that became a hot topic last September when it demonstrated Korea's first 'text prompt-based' music creation AI 'Mix Audio'.

As well as, Poja Labs was established in 2018 and has established itself as a mainstay of domestic music tech. Last January, they showed off their technological prowess by launching 'LAIVE', a music creation platform that may create a mix of composition, lyrics, and vocals inside 5 minutes.

The 2 firms recognize one another as rivals within the ‘domestic music creation AI industry.’ He also identified that the newly added Suno has outstanding benefits.

Poja Labs said, “Suno's characteristic is that it creates music with a transparent progression,” adding, “You may compose various song formats comparable to the primary verse, second verse, and chorus by entering the prompt, in addition to various vocals, sound effects, and instrument tones.” “Do it,” he said.

The evaluation is that usability facets comparable to the interface (UI) can’t be ignored. He explained that it’s convenient because you may enter prompts, view currently created sound sources, and past sound sources all inside one page. It’s also said that creating song titles and album covers is a novel feature.

Technically, it’s explained that it’s the ‘waveform generation’ method that’s currently most generally utilized by global tech. The reason is that Suno, in addition to Stability AI's 'Stable Audio' and Google's 'Refusion', recognize music as 'a single waveform' moderately than a mix of elements.

Nevertheless, problems were also identified.

A Newtune official said, “Suno seems to have placed priority on securing a position moderately than legal issues,” and added, “It seems that they’ve crawled a considerable amount of unlicensed data.” Consequently of monitoring the open chat room where Suno reviews were shared, it was said that 'circumstances of plagiarism' were also discovered.

For instance, if you happen to input the 'Korean ballad genre' with a big training data size, there may be a high probability of manufacturing an appropriately mixed result. Nevertheless, if you happen to input something like 'Korean Classic' with a small amount of existing data, the outcomes will inevitably be limited.

This can also be a limitation of AI created based on learning data. It was identified that because the song was created based on little data, direct problems arose, comparable to using lyrics that were greater than 5 syllables similar to existing songs.

Poja Labs also has an identical opinion. “There are cases where lyrics are so strange that it’s difficult to call them poetic permissions, or the prompts are ignored altogether,” he said. “It’s more vulnerable in areas where data has not been learned well.”

In actual fact, if you happen to input 'Please make Korean traditional music', the reason is that you just didn’t understand. Editing the sound source can also be unattainable. The evaluation is that the song selection itself seems to have been increased moderately than editing options. It is because two results are produced with one prompt input.


Meanwhile, he emphasized that domestic technology is at least that of Suno. Above all, Newtune has a multimodal creation function that goes beyond text prompts and inputs images to create background music. It’s also explained that the copyright issue was resolved by directly playing the instruments and beats utilized in learning.

Consequently, he emphasized that not only was the product of top quality, however the copyright issue was also resolved. It is alleged to be optimized for interactive music and simple to use to entertainment and metaverse. In an identical way, I listened to overseas ‘Endel’ and ‘Aimi’.

Forza Labs took a special approach called ‘MIDI creation’. It is simple to grasp if you happen to consider AI drawing electronic music scores. Particularly, Poja Labs also uses 100% self-constructed music data. Real-time editing of additional elements comparable to lyrics can also be possible.

(Photo = Poja Labs)

Moreover, in Korea, along with Newtune and Forja Labs, there are numerous music tech firms comparable to Creative Mind and Juice. Upon closer evaluation, all firms are adopting fundamentally different approaches to technology.

Ultimately, Newtune predicted that as a substitute of 1 company like Suno monopolizing the market, users would use quite a lot of tools to suit their tastes and purposes.

A Newtune official said, “Demand for generated AI can also be diversifying, comparable to meme-type music that’s popular on social media, interactive music that changes in real time depending on the listener's situation, and performance-type music that goes beyond listening to 'twiddling with' music.”

He also summarized, “This can be a trend that music tech and music creation AI firms don’t have any alternative but to diversify accordingly.”

Reporter Jang Se-min semim99@aitimes.com


What are your thoughts on this topic?
Let us know in the comments below.

0 0 votes
Article Rating
Inline Feedbacks
View all comments

Share this article

Recent posts

Would love your thoughts, please comment.x