The future of sound: Harnessing AI for audio branding

Artificial intelligence (AI) has not only become a buzzword over the past year but has also permeated various fields, including creativity. AI is now capable of generating texts, images, realistic videos, and even composing music. This development has elicited mixed reactions from musicians. Recently, many international artists signed an open letter advocating for the protection of authors’ rights against AI. However, some musicians embrace AI, declaring that it marks the beginning of a new era.

Before the AI era

Audio branding, when developed in alignment with the brand’s DNA, is a powerful tool that can shape the desired business image and influence customer attention and behavior. By eliciting emotional reactions, musical accompaniment can enhance a brand’s attractiveness and harmony. But how is this achieved? What steps should a company take to implement effective audio branding?

First, identify the type of music that will resonate with your audience. To do this, study your target audience and create a customer profile. Next, determine your brand values – what emotions should the music evoke, and what behavior should it encourage to achieve your business goals? This stage involves experimentation: use focus groups and test different genres and tempos of music in various settings. Once the musical style is established, select specific tracks either independently or with the help of professionals, and create playlists. It’s advisable to collaborate with industry experts, often streaming services for businesses that offer comprehensive audio design services, to ensure legal compliance and access to a vast library of licensed music.

AI composer has entered the chat

Many who experimented with generating images using AI have also ventured into creating musical compositions. There are open services available for this. For background music on social media, Amper and AIVA are popular choices. If you need to create a song, MuseNet can assist with melodies, chords, and even lyrics. Additionally, for style transfer – like transforming a favorite children’s song into an electronic version – Google Magenta is a suitable tool.

However, creating music “just to see how it works” isn’t the right approach for audio branding. While it’s true that an AI composer can be utilized, the “just try” method often results in monotonous and uninspiring music that won’t meet the ambitious goals set by marketers or business owners. Still, we shouldn’t be quick to dismiss this tool entirely.

As regards IT, AI tools can be used for three major groups of audio branding tasks:

automated tagging of the existing music database to determine parameters of each track;
expanding the music collection through adding generative pieces of music;
generating music streaming based on the company’s identity.

As regards marketing and business in general, the tasks for AI are obvious as well. Business is seeking to and already can do the following:

save resources, such as through creating small musical forms including commercial music, beats, ad jingles, and others;
avoid excessive royalty payments;
expand the database of music tracks in styles required by the brand;
create unique soundscapes (unique music pieces) for your particular environment.

Generative music pros and cons

Pros:

limitless creativity and a wide scope of musical ideas;
the ability to change the mood, genre, tempo and the instruments used;
producing music quickly and promptly, which is essential for experimentation;
production techniques that are more accessible than those used in traditional music writing.

Cons:

AI-generated music may lack emotional depth and nuance;
AI algorithms may produce music that seems generic;
AI performance is affected by the training data quality;
AI may have difficulty with complex forms and musical storytelling.

Ways to act: a case study

A fitness training studio approached us for background sounds and music for their classes. We have enough music for various types of fitness activities in our library, so we made a selection. The only difficulty was picking music for ballet classes as the studio sought to use some classical compositions.

There are not many classical pieces that can be used for ballet, and even fewer suitable for educational classes to teach tempo and rhythm. There is neoclassical music that is more new-age; there is not much of it, which often makes it difficult to contact its authors for permission to use their music. With few tracks available, students and teachers have to listen to the same piece at every lesson, which offers little room for a positive attitude and excitement. Some major schools try to diversify the experience by getting a musician to play live at the training sessions. Instead of selecting pieces, we chose to create our own music for ballet classes.

We developed detailed requirements for the musical form and sound of generative tracks, or neurotracks. A musical piece should have four beats per bar to serve for exercise repetition, be in 2/4 and 4/4 time signatures, and have a specific rhythm pattern and timing. The music was based on piano sounds, with a preset volume and spatial processing.

As the neural network is unable to get music compositions evolve and make tracks more appealing, we combined the generation parameters with those of different styles while maintaining the right balance. The AI needs to be ‘shown’ all possible musical patterns, with stylistic tokens used at the generation stage to get tracks in the desired style. Currently, we are testing the generated tracks and collecting feedback from the studio coaches, who will soon use music that sounds both classical and new.

Copyright

There is still no unified approach to regulating AI technology across the globe due to existing issues that lack an obvious solution. The essential question is who should be regarded as an actual author of a musical piece or score created with the use of AI from a legal perspective. The issue is not directly regulated by the Russian law, with very little case history.

The process of creating a music piece normally requires certain effort by a considerable number of people, and it must be legally protected. However, for now, the ownership of a newly created work depends on the contents of the contract between those who used the AI to create it and AI software developers, each case being specific. The neural network can be regarded as a processing tool rather than a method for creating music and audio recordings. If you choose to use AI in audio branding, you should rely on specific agreements and follow legislation and law enforcement practice news.