HANCE - Realtime AI audio enhancement for developers

Where Great Audio Begins

Every product developer has been there. You've built what you believe is a stellar app—intuitive interface, smooth user experience, powerful features—but then the reviews start rolling in. "Audio quality is terrible during calls." "Can't hear anything over the background noise." "The voice sounds robotic." Suddenly, your masterpiece feels incomplete, and your users are frustrated.

We know this problem intimately because we've spent decades on the other side of the microphone. Our team has spent over 100 years collectively wrestling with audio challenges—building plugins, developing signal processing algorithms, creating tools that thousands of audio professionals rely on daily. We saw the same pain points repeating across industries: developers who needed powerful audio enhancement but couldn't find solutions that were fast enough, private enough, or lightweight enough to embed in their products.

That's why we built HANCE.

We are HANCE, a deep tech AI company headquartered in Oslo, Norway, and we specialize in one thing: real-time audio enhancement through machine learning. Our technology addresses the core challenges that keep product developers up at night—noise that won't go away, delays that ruin real-time conversations, and privacy concerns that make cloud processing risky.

Our journey began in 2019 at the NAMM Convention, where engineers from Acon Digital and Soundly first met and discovered a shared frustration with existing audio processing solutions. The technology existed, but it was either too heavy for real applications, too slow for live use, or required sending sensitive audio data to the cloud. In 2021, we formalised our partnership and founded HANCE AS with a clear mission: to make professional-grade audio enhancement accessible, fast, and private.

Today, our technology powers applications across music production, podcasting, video conferencing, live streaming, and beyond. We've established partnerships with industry leaders including Intel, Nvidia, and Riedel, and our work has been showcased by renowned technology reviewer MKBHD on his YouTube channel. Most importantly, we've built a solution that respects both the privacy of your users and the computational limitations of your hardware.

---

What HANCE Can Do

We didn't set out to build another audio plugin. We set out to solve real problems that developers face every day—problems we witnessed firsthand from our partners in the audio industry. Each feature in the HANCE Audio Engine exists because a developer told us they needed it.

Real-time Stem Separation changed how music producers work. Imagine being able to extract the drums from a mixed track instantly, or isolate vocals for a remix, all without sacrificing the seamless flow of your production workflow. That's what our HANCE Audio Engine v2.0.1 delivers. Whether you're building a music app, creating sampling tools, or enabling new forms of creative collaboration, stem separation now happens in real-time—no waiting for renders, no quality loss.

For content creators, our Noise Removal technology means the difference between a professional-sounding podcast and one that sounds like it was recorded in a parking garage. The AI-driven deep learning model removes background noise while preserving voice clarity, and it does so with a model size of just 253 kB. That's remarkably small—we've engineered our technology to run efficiently on everything from powerful workstations to mobile devices.

When voice quality matters most, HANCE 26, our latest speech enhancement model, rises to the challenge. We optimized it specifically for the messy, unpredictable audio environments where real people actually make calls: busy offices, coffee shops, homes with background music. The model intelligently distinguishes between the voice you want to keep and the noise you want to remove, delivering clear audio without the artifacts that plague traditional filters.

Our collaboration with Acon Digital produced Extract:Dialogue, a tool that's become essential for video professionals. It pulls clean dialogue from complex audio scenes—music, sound effects, environmental noise—everything gets separated so you can achieve the professional mixing balance that makes films and videos engaging. This technology was featured by MKBHD on his YouTube channel, demonstrating how even challenging dialogue recordings can be salvaged.

What makes all of this possible is our unique approach: deep learning combined with traditional digital signal processing. This hybrid architecture gives us the intelligence to understand audio content while maintaining the precision that audio professionals demand. And because our models are trainable and customizable, they can be adapted to specific use cases—from specialised industry applications to unique product requirements.

Finally, our cross-platform audio engine is designed to go wherever your product goes. Built as a C++ library, it integrates into mobile applications, desktop software, and embedded systems. The same powerful technology that runs on a studio workstation can run on a modest processor in a consumer device.

---

Who Uses HANCE

We built HANCE for developers who need audio enhancement that works in the real world—not in ideal laboratory conditions. Here's how different users benefit from our technology.

Music producers and remix artists use our real-time stem separation to transform their workflow. Instead of spending hours manually separating audio elements or being limited to pre-separated stems, they can now extract drums, bass, and vocals instantly during production. One user told us they created an entire remix in a fraction of the time it would have taken with traditional methods—more time for experimentation, less time on tedious processing.

Podcasters and content creators rely on our noise removal and speech enhancement to deliver professional-quality audio without professional-grade recording environments. Whether they're recording in a home office with a noisy air conditioner or conducting interviews in busy locations, HANCE handles the audio cleanup in real-time. The result: cleaner productions with significantly less post-production time.

Video production professionals use Extract:Dialogue to solve one of the most common post-production challenges: dialogue obscured by background music or environmental sounds. Instead of forcing actors to re-record lines or using crude filters that damage audio quality, they can now extract clean dialogue and rebalance the mix. This has become particularly valuable for content creators working with interview footage, documentary material, or run-and-gun shooting scenarios.

Communication app developers integrate the HANCE Audio Engine SDK to enhance call quality in their applications. Low latency is critical here—we designed our technology specifically for real-time processing, ensuring that voice calls remain natural and responsive while unwanted noise disappears. Several video conferencing and calling applications have embedded our technology to differentiate their products in a crowded market.

Hardware manufacturers appreciate our ultra-lightweight models. At just 253 kB, our processing models run efficiently on modest processors, making them suitable for consumer electronics, IoT devices, and embedded systems that lack the computational power of dedicated computers. This opens possibilities for noise reduction and audio enhancement in hardware categories where such features were previously impractical.

Live streamers and broadcasters use our real-time audio enhancement to deliver better experiences to their audiences. Whether it's removing background noise from a gaming stream or ensuring clear voice audio during a live presentation, the ultra-low latency processing ensures nothing disrupts the viewing experience.

---

The Technology Behind HANCE

We could tell you that our technology is powerful, but we'd rather explain what that actually means for your product and your users.

Ultra-low latency isn't just a marketing claim—it's an engineering requirement we built around from the start. Real-time audio processing means processing audio as it arrives, with delays so small that users never notice. For live calls, streaming, and interactive applications, every millisecond matters. We've optimized our entire pipeline to minimise delay without sacrificing quality.

Privacy-first architecture reflects our belief that audio data belongs to your users. Our processing can happen entirely on-device, meaning sensitive audio never leaves the user's phone or computer. This is crucial for applications in healthcare, legal, finance, and any context where data privacy matters. When you integrate HANCE, you're not compromising your users' privacy to achieve better audio.

The 253 kB model size still surprises people when they first hear it. That's smaller than many images you'd find on a website, yet it contains sophisticated AI capable of understanding and enhancing audio. We achieved this through years of optimisation work, combining the best of deep learning with efficient signal processing. The result is technology that runs on processors that would struggle with much simpler audio tasks.

Our trainable and customisable approach means you're not stuck with a one-size-fits-all solution. Using our deep learning combined with DSP framework, you can train models on your own data for specific use cases. This has been valuable for companies serving specialised industries or those building unique products that require audio processing tailored to their exact requirements.

The cross-platform compatibility ensures HANCE integrates into your development stack regardless of your platform of choice. Our C++ library forms the core, with Python wrappers for data science workflows and HTTP APIs for cloud-based processing. Whether you're building for iOS, Android, Windows, Linux, or embedded systems, we have integration options that fit.

Behind all of this stands our technical team, with over 100 years of combined experience in audio processing. Our co-founder Stian Aagedal brings over 30 years of signal processing expertise from his work building Acon Digital. Our CTO Erlend Aune holds a PhD in Statistics from NTNU and previously led machine learning teams at Exabel and Amedia. Together with our team of engineers, we've built technology that draws on deep expertise across audio engineering, machine learning, and software development.

---

Ecosystem and Integration

HANCE doesn't exist in isolation. We've built our technology to work within broader ecosystems, partnering with companies and organisations that share our vision for better audio.

Our most significant partnership is with Acon Digital, the company behind our co-founders' original work. Together, we've developed Extract:Dialogue and Remix, audio plugins that have found homes in professional studios worldwide. These products demonstrate what's possible when our AI engine meets professional audio workflows.

We also work closely with Soundly, one of our founding companies, which provides AI-powered audio search and management tools. This relationship gives us deep insight into how audio professionals work and ensures our technology addresses real needs in content creation workflows.

Our relationships with Intel and Nvidia ensure that HANCE runs optimally on the processors and graphics cards that power millions of devices. We've worked with these platform partners to optimise our models, ensuring you get the best performance whether your product runs on Intel's integrated graphics or Nvidia's dedicated GPUs.

Riedel, a leader in broadcast and professional audio infrastructure, represents our reach into live production environments. Their confidence in our technology validates that HANCE meets the demanding requirements of professional broadcast applications.

We've received support from Innovation Norway and Nora AI, organisations that recognise the potential of Norwegian technology companies to compete on the global stage. Their backing helped us grow from a concept formed at NAMM to a company serving customers worldwide.

For developers ready to integrate, we provide C++, Python, and HTTP APIs to suit different architectures and workflows. Our GitHub repository offers open-source resources to help you get started, and our documentation covers everything from basic integration to advanced customisation.

Our technology serves industries ranging from music production and podcast recording to government, non-profit, and academic applications. Wherever audio quality matters and privacy is valued, HANCE can add value.

---

Frequently Asked Questions