Introduction
Krisp SDK for native desktop applications is a C-library.
Krisp SDKs are deployed and in use globally across more than 200M end-user applications and devices. Krisp SDKs power Krisp's own application in use by enterprises around the world, it also powers voice quality for Discord, RingCentral, Twilio and others.
The C-library is designed to be integrated into native apps running on Windows, Mac and Linux.
The library can also be integrated directly into Electron or other frameworks. Check the Integrations section for more details.
Supported Audio Characteristics
Audio frame duration
10ms
*
* While an audio stream of any frame size is supported, Krisp SDK processing is performed using 10ms frames.
Audio sample rates
8KHz, 16KHz, 32KHz, 24KHz, 44.1KHz, 48KHz, 88.2KHz, 96KHz
Performance
Reference Devices
![](https://files.readme.io/7021aa8-image.png)
Outbound (Mic) Noise Cancellation
![Benchmarks conducted for Desktop SDK v7.0.1 by processing 10ms frames.
CPU % = (Processing Time / Audio Length) \* 100](https://files.readme.io/0468678-NativeSDKv701.png)
Benchmarks conducted for Desktop SDK v7.0.1 by processing 10ms frames.
CPU % = (Processing Time / Audio Length) * 100
Inbound (Speaker) Noise Cancellation
![Benchmarks conducted for JS SDK v1.0.12 with Chrome v113 on Mac M1 and Chrome v114 on Windows and Mac Intel on the same reference computers used for Desktop.](https://files.readme.io/aaf0f3c-Native_SDK_v701_Inbound.png)
Benchmarks conducted for Desktop SDK v7.0.1 by processing 10ms frames.
CPU % = (Processing Time / Audio Length) * 100. Frame processing time is 0.5-2ms depending on platform. Package sizes and Memory footprint is measured for 8kHz model. 16kHz model adds ~10MB to both.
Background Voice Cancellation
![Benchmarks conducted for JS SDK v1.1.0 with Chrome v113 on Mac M1 and Chrome v114 on Windows and Mac Intel on the same reference computers used for Desktop.](https://files.readme.io/6a1ca24-Native_SDK_v701_BVC.png)
Benchmarks conducted for Desktop SDK v7.0.1 by processing 10ms frames.
CPU % = (Processing Time / Audio Length) * 100. Frame processing time is 2-4ms depending on platform.
Updated 6 months ago