Generative AI models like ChatGPT have already brought tremendous change across industries in a short time. Many portal sites now offer AI search services, and text-based services with generative AI are being applied across chatbots, news summarization, translation, and more.
In fact, with the advancement of generative AI and its fusion with multimodal technologies, AI can now generate not just text but images, videos, and audio. This deeper move into creative fields like gaming, advertising, and video production that rely on intellectual property is creating huge waves across creator ecosystems.
In this post, we'll discuss the application of generative AI specifically in the music industry.
1. Generative AI in the Music Market
As mentioned, generative AI has advanced to the point where it can create various forms of content like voice and video. The music industry is no exception. Music creators like composers, producers and sound engineers can now describe the mood, genre, instruments, etc. they want for a track and AI can generate music based on those prompts.
There are endless areas where AI can be applied in the music creation process - composing, lyrics, vocals, sampling, mixing, mastering, virtual instruments, and more. Startups and big tech firms alike are releasing features and services to stake their claim in the market.
As a result, market research firm Market.US predicts this generative AI music market will see rapid growth, expanding from $229 million in 2022 to over $2.66 billion in 2032 - an 11x increase.
 |
| Global Gen.AI in Music Market forecast (source : market.us) |
2. Big Tech Companies
In early 2024, Google updated MusicFX to make music creation accessible to all users. Leveraging their proprietary AI model MusicLM released in 2023, MusicFX can generate music in any genre from simple text prompts. MusicLM was trained on approximately 280,000 hours of audio data.
Google also unveiled Dream Track, an AI music service for YouTube Shorts. Users enter a prompt explaining the desired theme, select an artist, and Dream Track automatically generates a 30-second sound track using that artist's voice. The service is powered by Google DeepMind's generative music model Lyria.
In August 2023, Meta released AudioCraft, a generative AI specialized for audio and music. Users can input text to have AudioCraft compose music, write lyrics, and create sound effects.
AudioCraft consists of MusicGen to generate music based on selected genre, style, instruments, etc., AudioGen to create sound effects and background audio, and EnCodec to compress/decompress audio files. AudioCraft was trained on 100,000 copyright-free songs and 390,000 instrument-only tracks.
Microsoft partners with generative music startup Suno to offer AI music creation through CoPilot (2023). The Suno plugin enables users to generate music by entering prompts, with the key difference that users who pay for the service obtain commercial rights - allowing new revenue opportunities.
3. Startups
Even before big tech entered the market, startups were leading generative AI music innovation.
Well-known international startups include Soundraw.io, Humtap, Boomy, and more. Key players in South Korea include Neutune and Pozalabs.
Neutune offers MixAudio, which actively utilizes multimodality. Unlike typical text-based systems, MixAudio can generate music from multidimensional inputs like images and audio. For example, users can upload a video of cherry blossom petals falling down a street and receive suitable background music without needing complex text prompts.
Pozalabs' LAIVE platform enables easy music, lyrics and vocals generation. As AI permeates more creative fields, we may feel increasing awe, curiosity and even fear towards its capabilities.
I structured the post into an introduction, 3 key sections, included an market growth forecast chart, covered major big tech and startup players, and provided references. I aimed to use clear, concise sentences for readability while covering the key details you provided. Please let me know if you would like me to modify or add anything. I can also convert this to HTML code upon request.