Welcome to Lab99
Where the magic happens
Advanced audiology meets deep neural networks
RESULT: a revolution in voice isolation
At LAB99, audiology, software engineering, and deep learning converge to push the boundaries of what’s possible in extracting voices in even extremely noisy environments
LAB99 is our powerhouse—a unique innovation hub where advanced audiology meets AI inspired by the human brain. Just as the brain learns to focus on a single voice in a crowded room, our system has been trained on millions of voice samples to do the same—only faster, more precise, and more reliable.
Instant voice isolation
Once a voice is detected, our system instantly isolates and amplifies it, while intelligently suppressing surrounding noise. Think of it as the brain’s natural sound-filtering ability—supercharged by cutting-edge algorithms.
The concept may sound simple, but the science behind it is anything but. Years of deep research, advanced signal processing, and neural network modeling have led to a solution unlike any conventional noise reduction.
AI alone is not enough
A deep auditory understanding is required to truly reap the benefits
Learns just like you (did)
Highly complex learning from deep neural networks
Five years of dedicated deep learning research and millions of voice samples fuel our neural networks, enabling unparalleled precision in speech isolation. It is the same way a baby learns to understand voice, step by step, by listening several times to voices who have to be recognized. Learning the details, frequencies, tone-of-voice – so at the end of the day it can isolate e.g. the others voice form any other – even in extreme noise.
LAB99 isn’t just where we develop technology
Advanced Audiological Insight
Built with deep audiological expertise, Auto-SHARPI automatically adapts to unpredictable noise environments while preserving natural situational awareness.
Real-Time Extraction (<10ms Latency)
Our system processes sound in under 10 milliseconds, delivering clear, intelligible speech in even the most extreme noise conditions. Even this is groundbreaking, we are working on having even better latency time.
Military-Grade Reliability
Engineered to meet stringent military standards for performance, stability, and durability in the toughest operational scenarios. We are working on meeting even stronger standards than the military has.
Fits everywhere
Our solution is fully self-contained and can be deployed across diverse hardware and software environments, including standalone deep learning libraries.
— it’s where we invent the future of voice communication
How Sharpi works
Without data you’re just guessing…
How we take Voice Isolation far beyond Noise Reduction
Traditional noise reduction struggles in real-world environments, often creating “muffled” or “barrel-like” audio. Our voice-isolation approach leverages data-driven algorithms that separate speech from noise at the source, improving the signal-to-noise ratio with unprecedented accuracy. Inspired by CNN-based models like AlexNet, we pioneered a voice separation model in 2021—laying the foundation for SHARPI, our breakthrough technology.
Prepare to be impressed
Trained on 25 years of audio
Decades of real voices teach the system to catch every nuance, so it understands speech the way people naturally speak.
200 mio. operations pr. sec.
In a single second, the network runs hundreds of millions of tiny checks to keep up with the flow of natural speech.
4 mio. weights adjusted
Millions of tiny dials are fine-tuned inside the system, shaping how it listens and improving accuracy with every decision.
Reduced to math
Speech becomes pure numbers, allowing the network to run complex tasks efficiently inside closed systems with very limited processing power.