Overview: The Future of Digital Music

ACE Studio is an AI-native music workstation designed for producers, composers, and professional songwriters. The company provides cutting-edge tools, including vocal synthesis and AI-driven instruments, to streamline the creative workflow. As CTO and Co-founder Sean Zhao describes it, “ACE Studio is a next-generation music creation tool for producers and composers”.

While the platform initially launched for professionals creating AI Singing Voice Generation, it has expanded to include melody generation and layering features that allow users of all skill levels to create music without needing deep professional expertise. When ACE Studio faced challenges with latency,cost, and the complex setups of hyperscalers, they turned to DigitalOcean and AMD for their GPU and inferencing needs, which resulted in lower costs, fewer lags, and a better experience for their customers.

The Problem: High Latency and Prohibitive Costs

As ACE Studio’s user base grew and their models evolved, they faced a significant two-fold infrastructure challenge involving both training and inference. They found that popular GPU types from major cloud providers including AWS and Google Cloud were often expensive and suffered from limited availability, which hindered their ability to scale. At the same time, local low-cost data centers proved to be unreliable. As CTO and co-founder Sean Zhao said “Sometimes we need to scale up, but hyperscalers don’t have any GPUs left. The low-cost local data centers have fragile reliability and stability. We were looking for a provider like DigitalOcean that sat between these two options, which had availability but also reliability at scale.”

These infrastructure limitations resulted in high latency, which was a deal-breaker for creators who require an ultra-low latency experience when working with MIDI and lyrics. Users experienced latency of over two seconds, an unacceptable delay when professional producers and hobbyists make frequent, small adjustments to a song. Additionally, the high costs and limited availability of hyperscale providers restricted how often ACE Studio could run training models, which was a bottleneck for the team as they tried to stay ahead of competitors.

The Solution: High-Performance Infrastructure with DigitalOcean and AMD

To solve these challenges, ACE Studio migrated to DigitalOcean GPU Droplets powered by AMD Instinct™ MI325X GPUs.

For ACE Studio, the move to DigitalOcean meant a significant upgrade in both their operational experience and overall GPU performance. CTO Sean Zhao highlighted that the DigitalOcean platform was a welcome change from hyperscalers, saying that “DigitalOcean is easy to use and has a simple user interface which my developers like, especially compared to traditional hyperscalers, which have been a nightmare for them.”

This ease of use translated directly into minimal operational overhead for his team in setting up their models. Beyond simplicity, the overall solution delivered on their needs for better reliability. This was a crucial factor for ACE Studio after dealing with reliability challenges from low-cost local data centers, and ensures their high-speed creation tools maintain continuous uptime for their professional users.

This combination of DigitalOcean’s simplified management and AMD’s high-performance GPUs provided a massive boost to their development speed, and training cycles that previously took half a month were reduced to less than one week. As Zhao explained “We can see the result very quickly and we can evolve very quickly. It is crucial for us to keep advancing in the market to satisfy our users and not be overtaken by competitors.”

ACE Studio’s professional creators require near-instantaneous response for their work and have a low tolerance for latency. Specifically, when they incorporate elements like MIDI melodies and lyrics, they demand almost instant feedback. To meet this critical need, ACE Studio implemented a two-part latency reduction strategy.

To address their geographically diverse user base—spanning Europe, North America, and East Asia—they deployed inference machines strategically in these zones to minimize network latency.

This was all achieved while maintaining a superior price-to-performance ratio compared to all other solutions ACE Studio had evaluated. The combination of DigitalOcean and AMD provides ACE Studio with a strong competitive advantage. Since adopting DigitalOcean GPU Droplets with AMD Instinct GPUs, ACE Studio has seen measurable improvements across both inference and training workflows:*

These combined efforts resulted in some pretty impressive achievements:

• Improved infrastructure cost efficiency by about 5%

• Inference latency reduced from 2500ms to 1500ms - a 40% decrease

• Model training cycles cut from ~2 weeks to under 1 week, doubling iteration speed

• Improved cost efficiency, with a strong price-to-performance ratio from AMD Instinct GPUs on DigitalOcean, enabling more experimentation within the same budget

• Reduced operational overhead due to faster GPU provisioning and simplified management

• Increased reliability which eliminated the downtime risk that had plagued their previous low-cost data center setup

Looking Forward: Scaling the Creative Barrier

With a stable and cost-effective infrastructure now in place, ACE Studio is prepared for a major surge in users as AI technology lowers the barrier for human masterpiece creation. Zhao emphasizes that this technology is a tool meant to supercharge creators rather than replace them, as true inspiration and ideas still come from the human mind.

Looking ahead, the company plans to deepen its partnership with DigitalOcean and AMD to support increasingly large music generation models. A key focus for their future roadmap is the implementation of a high-performance distributed file system to handle the massive amounts of data required for audio model training, ensuring they maintain their leadership position in the rapidly evolving AI music market.

"Access to large-memory, cost-efficient GPUs expands the scale and complexity of the models we can build. The infrastructure provided by DigitalOcean and AMD is a key enabler for our roadmap, allowing us to iterate faster and deliver new capabilities to creators.” — Sean Zhao, CTO & Co-Founder, ACE Studio

*Results based on specific customer configurations and may vary.