TikTok has reportedly started testing an ‘AI Song’ feature, allowing users to auto-generate music via on-platform text prompts. Photo Credit: Jonathan Kemper
TikTok has reportedly become the latest platform to embrace artificial intelligence music, as some of the app’s users can now auto-generate custom tracks with text prompts. Observant Redditors spotted and then posted about the feature at hand, dubbed AI Song and seemingly available to select TikTokers in the service’s video-upload section. According to the appropriate on-screen description, the artificial intelligence offering is powered (at least in part) by large language model (LLM) Bloom.
Per its own description, Bloom boasts 176 billion parameters and can “generate text in 46 natural languages” – a significant point given TikTok’s global userbase. At present, those with access to the test – it’s unclear precisely how large this group is or when a broader rollout could arrive – can make music for their TikTok clips via text inputs, as mentioned.
Bearing in mind the customization flexibility and cost advantages of AI music – notwithstanding its potentially devastating long-term impact on actual musicians – logic suggests that an AI Song expansion will come to fruition sooner rather than later.
Of course, the artificial intelligence pivot, reportedly one component of a wider strategy at TikTok, didn’t happen overnight; the app and its ByteDance parent have been testing the AI waters for a while now. To be sure, Beijing-based ByteDance (via a subsidiary named Funnico) released an AI “music creation tool” called Ripple last summer. As described by the relevant App Store summary, the standalone service enables one to generate lyrics and vocals, extract components of audio uploads, and more.
Bigger picture, it stands to reason that Ripple is already part of TikTok’s AI plans and will eventually factor prominently into the short-form mainstay and/or TikTok Music. More immediately – and in spite of this apparent openness to AI in certain areas – TikTok only recently announced a purported objective of “protecting election integrity in 2024.”
As one component of said objective, head of USDS trust and safety Suzy Loftus (a former San Francisco prosecutor who reportedly worked for now-Vice President Kamala Harris) indicated that TikTok will tackle “misleading AI-generated content” in the political sphere.
Per Loftus, that effort will include continuing to block “manipulated content that could be misleading, including AIGC [AI-generated content] of public figures if it depicts them endorsing a political view.”
Additionally, TikTok creators must “label any realistic AIGC,” according to Loftus, whose employer has faced criticism over alleged on-platform misinformation, besides high-profile lawsuits for allegedly exposing minors to harmful content and multimillion-dollar fines related to the misuse of children’s data.
