top of page

CASE STUDY 02 / AUDIO NETWORK

The Power of AI in Music Searching

Our Music library is made up of 250,000+ tracks composed for TV, Film & Advertising. Our users find it time-consuming finding the right track for their project, and are often working towards very tight deadlines. Recently, we’ve introduced two new methods of searching our music library, both powered by AI - Audio Search and Prompt Search.

Competencies

UX, UI

Understanding AI Technologies 

Product Marketing

​

Role

Lead Product Designer

Observer in Research & User Testing

Prompt Search Forest.png
test.jpg

User Problem

Our users are video editors who are often pressed for time and working towards very tight deadlines to find the perfect track to help bring their story to life. 

Solution

Utilising AI, we implemented two new methods of searching our music library - Audio & Prompt Search. We were one of the first in our market to implement both.

Results

Both features have been hugely popular with our users, in particular with Television Editors. We found that Prompt Search has been extremely popular with non-native English-speaking users where our music data is not comprehensively localised. We've also learnt from speaking to users that some are still a bit dubious of AI - which is a great learning to take forward. We need to reassure users when and how we're using AI. 

Our Clients

Our clients are video editors working in Television, Film, Advertising or for a Brand, such as Lego, Disney and the BBC. Their priority is finding the right track(s) for their project. They often face very tight deadlines and work in low-light editing environments. 

Pain Points 

​Traditional keyword-based search can be limiting, requiring familiarity with specific terms from our taxonomy. This poses a significant hurdle, especially for non-native English-speaking users, given our most descriptive music data is not comprehensively localised.

Solution

Prompt Search addresses these issues by enabling users to search using intuitive and everyday language, eliminating the need for extensive music knowledge. For example, users can search for something that describes the scene they’re working on like: zebras roaming the Savannah or a magical underwater world. 

​

​We didn't want prompt searching to exist as a separate experience. Instead, we wanted it to be integrated seamlessly into our primary search tool that our users know and love. 

We used the universal 'AI' sparkle to indicate when a search was a "prompt" and therefore powered by AI. We also included a description in the UI to explain what a prompt search is, and how it differs from a regular keyword search (see above).

 

We debated what to call the feature. Initially we discussed "AI Search", but some users revealed they were still slightly dubious of AI technologies. We settled on "Prompt Search" as we were starting to see "Prompt" become a more popular search term. We ensured we provided plenty of examples of a "prompt" in both the Search UI and promotional material, for example "music for an uplifting spring wildlife documentary". 

EDUCATING USERS

We provided a support page to explain how the AI works and to reassure users how and when we're using AI. We also provided further education through the use of a tooltip to explain how you can refine your search, such as by adding references like places, movie characters or a time-period. ​

Prompt Tooltip.png

Next, we implemented Audio Search which allows users to paste a link to an existing track into the search bar to find similar-sounding results in our catalogue.

Pain Points

Our clients often have an existing commercial track in mind when editing a video project (for example, Dolly Parton, Working 9 to 5). However, clients can not always licence that track. Therefore, they need to find something similar. 

Solution

Users can now paste any link (e.g. YouTube, Spotify, TikTok etc.) into our search bar which will return similar-sounding tracks in our catalogue e.g. a track with a similar vibe, beat, or even similar lyrics. We know how important this type of innovation is to our users who frequently work under tight deadlines, allowing them to cut down their average search time by as much as 90%.

Audio Search 2.png

Promotions

I worked with the Marketing team to promote Audio and Prompt Search. We usually find that new product releases are hugely popular with users because these types of innovations can significantly improve their workflow, which tends to be fast-paced and stressful. 

​

We created support pages for both Audio and Prompt Search to explain how to use the features and to provide more information on how the AI works. We also promoted the features on our logged-in homepage in our 'what's new' carousel. Both features received the highest click-through-rates out of any content we've promoted in that position. This shows how important these types of workflow innovations are to users. 

Screenshot 2025-03-03 at 12.26.45.png
Screenshot 2025-03-03 at 12.27.15.png
Screenshot 2025-03-10 at 15.46.02.png

This graph shows the engagement of the Audio Search Promo banner on our logged in homepage (dark blue line), compared to other content in the same position. This confirms that these types of workflow innovations are very important to our users. 

Learnings so far

These two AI features are recent implementations on our site (2024), but so far both have been very popular with users, especially "high-value" television editors. We will continue to learn about how users are using these features, and iterate to ensure we offer the best search experience.
​
As expected, Prompt Search is popular with non-native English-speaking users where our music data is not comprehensively localised.

There are potentially other areas of our experience where both Audio Search and Prompt search AI could be utilised and beneficial to users. For example, when users submit a brief, or when they create a project.

In research, we found that some users are still slightly dubious of AI technologies, which is something we need to consider going forward, for example, reassuring users when communicating how we're using AI. 
bottom of page