Suno, a ChatGPT-Powered Chatbot, Can Generate AI Music Using Text Prompts
Suno is an artificial intelligence (AI) music generator that can create original songs using a simple text prompt. The AI platform was first unveiled in July 2023 when it moved to open beta and began adding users to its Discord channel to test the model. Later, it also allowed users to generate AI music on its web interface. In December 2023, the platform collaborated with Microsoft to add an extension for Copilot that allowed users of the Chatbot to generate songs as well.
On Sunday, a report by the Rolling Stones highlighted the deeper workings and the vision of the company. As per the report, Suno only generates original music based on simple text prompts. Once the prompt has been entered, the AI model generates a 15-second long song within a minute. The platform has both free and paid tiers — Pro and Premier. Among many upgrades, paid subscribers are also given general commercial rights to the song, while free account holders are forbidden from monetising the songs. At present, the cheaper Pro tier starts at $10 (roughly Rs. 830) a month.
We, at Gadgets 360, tested the web interface of Suno and found it to be user-friendly and easy to navigate through. Within seconds, we were able to type our prompt and generate original tracks that sounded unnervingly realistic. We used the prompt, “A punchy pop song about an AI stuck inside the screen, jealous of real human beings,” and in less than a minute it created two versions of the song titled ‘Digital Envy. The generated song can be heard here.
The report highlighted that while music creation is done by Suno’s native AI model, for lyrics and song titles, it uses OpenAI’s ChatGPT. The US-based startup has not revealed the architecture or any other details about its AI model. The secrecy also gives rise to a conflicting problem that the source of the data it was trained on is not known. This became an issue last year as the US music industry expressed concerns to regulators over AI models using their copyrighted songs to train on, and then create deepfaked songs that sound similar to their voice.
However, Suno navigates through the tricky situation by not generating any music that is sung in the style of a real artist. If a prompt like that is entered, it refuses to generate music. As per the report, the AI startup is also in communication with major music labels. On February 23, Suno publicly released its V3 Alpha model for Pro and Premier users, which the company claims “creates more realistic, authentic music” music. Only V1 and V2 models are available for free users.