ByteDance, the parent company of TikTok, has cemented its position as one of the most influential AI developers globally. The Chinese tech giant, valued at $300 billion, is aggressively investing in AI, including procuring billions of dollars worth of AI chips. Competing directly with OpenAI, ByteDance’s AI division is making waves, particularly with its AI assistant, Doubao, which boasts 78.6 million monthly active users as of January 2024—ranking just behind OpenAI’s ChatGPT.

One of ByteDance’s latest ventures is Seed-Music, a sophisticated AI music generation platform that promises to lower barriers to music creation. However, while its technological advancements are impressive, they also raise critical ethical and legal concerns regarding AI’s role in music composition.

Doubao’s Evolution into AI Music Generation

Doubao has evolved beyond a chatbot, integrating multimodal AI capabilities, including text, image, and audio processing. In September 2023, ByteDance introduced AI-powered music generation within the Doubao app, allowing users to compose music and lyrics in a matter of clicks. The expansion of this technology culminated in the launch of Seed-Music, an advanced AI system designed to empower both amateur and professional musicians.

How Seed-Music Works: The AI Pipeline

According to ByteDance’s research team, Seed-Music operates through a sophisticated AI architecture based on three core representations:

Audio Tokens – Converts input descriptions (e.g., lyrics, reference audio, or music style) into a structured audio sequence.
Symbolic Tokens – Provides an interpretable format, such as MIDI and MusicXML, which can be modified by musicians.
Vocoder Latents – Uses a deep learning model to generate high-quality audio output at 44.1kHz stereo resolution.

This AI-powered workflow enables users to generate complete songs with minimal manual input, making it particularly accessible to those without formal music training.

The Role of Music Information Retrieval (MIR) Models

A critical component of Seed-Music’s development involves Music Information Retrieval (MIR) models. ByteDance has built its own MIR tools to extract musical features such as beat tracking, key and chord detection, structural segmentation, and multi-instrument MIDI transcription.

This process allows AI to analyze existing music and generate new compositions with stylistic accuracy. However, the use of MIR models in training AI music systems brings up questions about data sourcing, especially regarding copyrighted works.

Controversy: Did ByteDance Train AI on Copyrighted Music?

ByteDance’s research team cites the Isophonics dataset, a collection of music annotations featuring artists like The Beatles, Michael Jackson, Queen, and Carole King. This dataset, originally compiled by Queen Mary University of London’s Centre for Digital Music (C4DM), has been widely used in MIR research. However, its application in a commercial AI music product raises concerns over whether AI-generated compositions could inadvertently replicate elements from copyrighted songs.

Although ByteDance has not explicitly confirmed whether its AI models were trained on copyrighted recordings, its research into song segmentation and structure analysis—using data from well-known artists—has sparked debate within the music industry.

Ethical and Legal Considerations

ByteDance asserts that Seed-Music is designed to assist rather than replace musicians. The company has outlined a commitment to ethical AI use, stating:

AI should support human creativity, not disrupt musicians’ livelihoods.
AI tools should offer neutrality and diversity in artistic expression.
Vocal synthesis technology should include safeguards against deepfake misuse.

ByteDance has implemented watermarking and multi-step verification to prevent unauthorized use of AI-generated vocal tracks, particularly in impersonating real artists. However, as with many generative AI models, the potential for misuse remains a concern, especially as AI-generated music becomes more sophisticated.

Potential Benefits of AI-Generated Music

Despite the controversies, Seed-Music presents several potential advantages for musicians and content creators:

Accessibility – Allows anyone to compose music without technical expertise.
Efficiency – Streamlines songwriting, reducing time spent on arrangement and production.
Innovation – Introduces new creative possibilities through AI-assisted composition.
Cost Reduction – Lowers production expenses for independent artists.

By leveraging AI, ByteDance aims to democratize music creation, enabling both professionals and novices to produce high-quality compositions with minimal effort.

Industry Implications and the Future of AI Music

The music industry is still grappling with the implications of AI-generated content. While companies like OpenAI, Google, and now ByteDance continue to push the boundaries of AI music, legal frameworks have yet to catch up.

Key concerns include:

Intellectual Property Rights – How should AI-generated music be copyrighted?
Artist Compensation – Should musicians whose styles influence AI compositions receive royalties?
AI Transparency – Should companies disclose what datasets their models were trained on?

With AI continuing to reshape content creation, the question remains: Will AI empower musicians, or will it disrupt the music industry as we know it? Only time—and regulation—will tell.

License This Article

Source: Music Business Worldwide

$1.00

$2.00

$5.00

$10.00