TikTok is currently testing a new feature called “AI Song,” which allows users to create songs using artificial intelligence based on prompts they provide, as confirmed.
This experimental feature is available only to a limited number of users and is intended for entertainment purposes, not for creating the next viral hit.
According to TikTok, the AI Song feature uses Bloom, a large language model, to generate lyrics from text prompts provided by users. These lyrics are then matched with music from a pre-existing catalog within the app.
TikTok user Jonah Manzano shared a walkthrough video demonstrating how the feature works. The process begins with AI Song prompting users to “enter a prompt to generate a song.”
TikTok offers suggested prompts like “Going to concerts,” “A new beginning,” and “Cuddling with pets,” though users can also input their own. Once a prompt is entered, the app generates a song.
After a brief moment, the app produces lyrics relevant to the prompt and pairs them with a musical track.
Users can adjust the song’s vibe by choosing between genres such as “pop,” “EDM,” and “hip-hop.” The end result is a unique song with lyrics that align with the entered prompt.
The video also notes that there are certain guidelines for using the AI Song feature.
For example, it cannot be used to exploit or harm minors, spread false or personal information that could be damaging to others, or engage in defamation, impersonation, or harassment.
As with other test features, it is unclear when or if AI Song will be widely released. TikTok mentioned that the feature’s name might be changed in the future.
TikTok is not the only platform exploring tools to help users create music for short-form videos. A few months ago, YouTube introduced a “Dream Track” feature that enables creators to produce 30-second AI-generated soundtracks.
Users can generate these soundtracks using the voices of artists like Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Sia, T-Pain, Troye Sivan, and Papoose.
While TikTok’s AI Song feature is not specifically geared toward the music industry, ByteDance, TikTok’s parent company, launched a free music production app in the U.S. last June called Ripple.
Ripple helps musicians create and edit audio, functioning similarly to portable digital audio workstations (DAWs).
Users can sing or hum a melody into Ripple, which then uses machine learning to expand the melody into an instrumental track.
The app’s model is trained on music that ByteDance either owns or has licensed. Ripple also allows users to record, cut, trim, and edit audio files directly on their mobile devices.
The introduction of the AI Song feature comes as TikTok continues to delve deeper into the music industry. Last year, it launched its own subscription-based music streaming service called “TikTok Music” in select countries, aiming to compete with platforms like Spotify and Apple Music.
Though TikTok has not disclosed a timeline for launching the streaming service in the United States, ByteDance filed a trademark application with the U.S. Patent and Trademark Office for “TikTok Music” over a year ago, suggesting that a U.S. launch may be forthcoming.
