📝 What is Krisp AI Voice SDK?

Krisp AI Voice SDK is a collection of ultra-low latency AI models that improve speech clarity in real-time communication applications such as

Voice AI Agents
Online video and audio meetings
Mobile calls
Streaming, broadcasting or podcasting

AI Voice SDK is deployed on more than 200M+ devices and is processing 75B+ mins of speech every month. It is integrated inside apps such as Discord, RingCentral, Synthflow, Vapi and 100 more.

The AI models included in SDK are extremely small and run on CPU.

There are two SDK packages for 2 distinct use cases:

VIVA SDK - designed for human-to-AI communication (Voice AI Agents)
RTC SDK - designed for human-to-human communication (Calls, Meetings)

Both SDKs include an impressive collection of AI technologies.

VIVA SDK models

Voice Isolation

Isolates the main speaker's voice in real-time. Fixes false-interruptions, Turn-Taking and improves WER (Word Error Rate).

Turn-Taking

Predicts when a user is likely to finish their turn, enabling Voice AI Agents to respond naturally, without awkward pauses or interruptions.

RTC SDK models

Built for real-time calls and meetings: noise cancellation, background-voice cancellation, accent conversion.

Accent Conversion

Converts users' accents while maintaining their voice, in real-time.

Noise Cancellation

Removes background noises, both inbound and outbound, in real-time.

Background Voice Cancellation

Removes background voices and noises from the main speaker's voice, in real-time.

How Does it Work?

All Krisp SDKs are deployed on-prem, customer-side.

Runs on-device or in cloud (on-prem)

Integrates on server with on-prem (Voice AI Agents)
Integrates into the local application’s audio path (e.g. WebRTC)
Runs as a module in frameworks such as Pipecat and Daily

Minimal latency and compute

Ultra-low latency for real-time communications
Runs on CPU

Multiple models for different needs

Supports narrow-band, wide-band and full-band audio
Supports different use cases (telephony, WebRTC, etc)

Getting started

Start by choosing your path:

Start with the VIVA SDK →
Start with the RTC SDK →

From there you'll get setup instructions, sample code, platform considerations, model guidance, and troubleshooting resources.

Happy building—let’s make clean, powerful voice experiences together!