Traditional methods for noise suppression work by observing the steady state energy in different frequency bands when you are not speaking, making the assumption that this is unwanted noise. The algorithm then subtracts that noise energy in each band to clean up the audio. The downside of this approach is that it mainly removes steady sources of noise, such as fans and air-conditioning, and is less effective at removing non-stationary noise sources, such as keyboard typing, mainly when you are speaking at the same time. This approach can also lead to distortion of your voice, especially when the noise is loud.
With the advent of AI-powered processing techniques, it is now possible to classify more precisely what components of audio are coming from your voice and what components are coming from background noise. This approach involves training a Deep Neural Network (DNN) with thousands of hours of labeled audio data to teach the network how to discriminate between voice and other sounds. This discriminatory power allows the noise suppressor to precisely remove unwanted components of audio, maintaining the quality of the target voice.
This technique allows the algorithm to handle a broader range of noise types, giving it a significant advantage in real-world noise suppression applications. It can also safely remove noise, even while you are speaking. This technique is not limited to eliminating noise; it can also be used to remove the reverberant part of speech that is captured when your microphone is not near your mouth.
The result? You hear less of the background noise and more of the conversation, leading to clearer, more effective communication, whether you’re in a busy cafe, a bustling office, or just dealing with noisy neighbors.
Our Competitive Edge
Superior Performance
Firstly, let’s talk about performance. We’ve developed a robust noise reduction mechanism that outperforms traditional techniques. Don’t just take our word for it – we have a video demonstrating our noise suppression in action. Watch as the distracting hum of everyday life disappears at the flick of a switch, leaving behind only clear, undisturbed communication.
Beyond just noise
Reverberation – the reflection of sound waves from surfaces in an environment – can often distort audio and make it harder to understand. It’s particularly noticeable in large rooms or spaces with hard surfaces, where sound can bounce around and create an echo-like effect.
Our new models available in Web SDK 3.11 and upcoming mobile SDK 3.11 releases, trained using sophisticated Deep Neural Network techniques, can accurately identify and reduce this reverberation.
Build for Real-Life Scenarios
While high-end devices with advanced processors can comfortably run complex DNN algorithms, lower-end devices might need help, leading to suboptimal user experiences. Recognizing this, we have designed our noise suppression technology to be dynamically scalable.
By striking a balance between noise suppression and device performance, our technology ensures a better holistic experience. Regardless of the device they’re using, users can enjoy clear, uninterrupted conversations without worrying about their device slowing down or their battery draining quickly.
Wide range of availability
At Dolby, we believe everyone deserves extraordinary audio, no matter the platform or use case, virtual worlds, video calls, live digital events, or gaming – we’ve got you covered. Dolby delivers superior noise suppression, enhancing your audio experience wherever you are, whatever you do.
The following table shows the feature availability across all our client SDKs.
| Product | Availability |
|---|---|
| Web SDK | ✅ starting 3.11 [*] |
| IOS | ✅ Since 3.7 |
| Android | ✅ Since 3.7 |
| Unreal Plugin | ✅ Since 1.0 |
| Unity Plugin | ✅ Since 1.1 |
| C++ SDK | ✅ Since 2.1 |
| React Native SDK | ✅ Since 3.7 |
| Flutter SDK | ✅ Since 3.7 |
Compact Model Size
Next, size matters. Our model is highly optimized. This makes it lightning-fast to load, particularly on the Web platform where every byte counts. With us, you’ll experience less lag, less waiting, and more productive collaboration.
Transparent Pricing
At Dolby, we believe that the best technology should be accessible to everyone without any hidden costs or caveats. We do not charge extra for our DNN noise suppression. We’re committed to delivering top-tier technology without locking you into a complicated pricing scheme.
APIs
At the heart of the noise reduction technology is the setCaptureMode API, which gives you the power to control how the noise reduction works according to your specific needs. Two main modes are available – Standard and Unprocessed. In Standard mode, you can configure the level of noise reduction.
To help you understand the differences, here’s a summary table:
| Mode | Audio Processing Technology | Ideal For |
|---|---|---|
| Standard – High | Cutting-edge DNN noise reduction, where the AI comes in to intelligently distinguish between speech and noise for superior audio clarity. The SDK automatically switches back to traditional noise reduction on unsupported platforms. | Voice communication with maximum noise reduction |
| Standard – Low | Traditional noise reduction, while not as advanced as the DNN approach, it can still help to reduce common types of background noise. | Voice communication with moderate noise reduction |
| Unprocessed | No additional optimization, the SDK lets the audio pass through without applying any optimization | Non-voice audio inputs or when natural ambient noise is desired |
const captureMode = {
mode: 'standard',
modeOptions: {
noiseReductionLevel: 'high', // See: https://docs.dolby.io/communications-apis/docs/js-client-sdk-model-audiocapturemodestandardoptions
},
};
await VoxeetSDK.audio.local.setCaptureMode(captureMode);With the setCaptureMode API, you’re in control. Whether you’re hosting a podcast or webinar, sharing a outdoor sports event, or conducting a business meeting, you can tailor the noise reduction to suit your unique audio environment and requirements.
Conclusion
To summarize, our AI-powered noise suppression technology brings an unparalleled advantage to the world of real-time communication. Its innate ability to intelligently differentiate between speech and noise assures clear, high-quality audio. Moreover, the algorithm’s capacity to adapt based on device capability ensures optimal performance without sacrificing user experience, even on lower-end devices. With universal availability across all platforms – Web, mobile, desktop, and gaming engines – we’re breaking down barriers to deliver superior noise suppression everywhere. The configurable setCaptureMode API offers flexibility, allowing users to customize their audio environment as needed. In essence, our technology is not just about suppressing noise, but about elevating communication. Wherever you communicate, let’s make every word count.
