The Journey to Seamless Background Music for Pixigon Users

A Collaborative Beginning: Crafting Music for Pixigon

To realise our vision, we knew we needed a top-tier musician and producer. That’s why we partnered with Johan “Kermit” Bobäck, a seasoned professional with experience in creating stock music for commercial productions. He brought a unique blend of creativity and technical expertise to the table, which was precisely what we needed. However, we had some very specific requirements.

The challenge was to create music that worked seamlessly with videos ranging from five seconds to several minutes. A tall order, but Kermit, who also happens to be a talented programmer, understood the complexities and executed the task with finesse.

Even though Johan’s work was efficient, bespoke music production comes at a significant cost. Thus, we began exploring ways to broaden Pixigon’s capabilities and make it easier — and more affordable — for our users to find the perfect background track.

The Expansion: Evaluating AI-driven Music Solutions

From the outset, Pixigon allowed users to upload their own background music. However, as AI-driven music services rapidly expanded, we saw an opportunity to evolve. The timing was perfect to explore what the market had to offer and potentially integrate a strong candidate directly into the Pixigon app.

Before diving into the market, we established a few essential criteria for any music solution we would consider:

• Royalty-free music: The tracks must be free of ongoing royalties to ensure users can utilise them without legal complications.
• Cost-effectiveness: The service must be affordable while maintaining quality.
• Precision: The solution should help users find a background track that complements the tone and message of their video.
• Integration-friendly API: For services offering integration, a robust API was a must to meet our technical standards.

With these criteria in mind, we did some research, identifying a shortlist of promising candidates in the market. It’s worth noting that this sector is evolving rapidly, and some services might change or disappear quickly. For example, one service listed in an article we used for reference had removed public access just 14 days after the article was published.

Putting AI-driven Music Services to the Test

To assess these services from a user’s perspective, we created a fictional marketing video and used each service to generate a background track. The aim was to evaluate how well the music matched the video’s content and message, simulating the typical Pixigon user experience.

As part of this process, we also verified if each service offered an API for potential integration into Pixigon. Depending on the interaction model of each service, we crafted a prompt designed to guide the AI to produce a background track that aligned with the mood and message of the video.

Here’s the general prompt we developed:

Create an upbeat, modern background track with a positive and professional mood. Use a medium-fast tempo with light electronic instruments, soft percussion, and subtle synths. The track should be dynamic, engaging, and suitable for a promotional marketing video.

The Results

One key takeaway from our tests was that, as with most prompt-driven AI services, crafting precise prompts is essential. This is something that improves with trial and error and through interaction with the model.

For Pixigon’s specific needs, we ultimately chose Soundraw.io and Beatoven.ai. Both services provided the API we required and allowed us to easily generate a track that suited our use case. While the other contenders were impressive, they lacked the integration features we needed, or were not as good.

Integration Possibilities

As we continue evaluating which service to integrate with Pixigon, the shortlist has narrowed down to two: Soundraw.io and Beatoven.ai. Stay tuned for our follow-up article, where we’ll reveal our final decision and offer further insights into the integration process.