SDK or API: choosing an integration

Krisp offers two product families. Both bring Krisp's voice AI into your product. They differ in where the audio is processed and in how you integrate them.

Krisp SDKs

The VIVA SDK and the RTC SDK are libraries that you integrate into your application, web app or server. Their AI models are small, run on CPU and process audio customer-side, inside your application's audio path. The VIVA SDK is designed for human-to-AI communication (Voice AI Agents). The RTC SDK is designed for human-to-human communication (calls and meetings).

Krisp API

The Krisp API gives you Krisp's speech models as realtime WebSocket endpoints in Krisp's cloud. You open a WebSocket connection to wss://streaming.krisp.ai/<endpoint>, configure the session with a JSON message, then stream audio or text and receive results on the same connection. There is no library or model file to ship with your application.

Comparison

	Krisp SDKs	Krisp API
Where processing runs	Customer-side: on the device or on your own servers. The Voice Translation SDK is the exception: it streams audio to Krisp Cloud.	Krisp's cloud, currently in the US region. Clients outside the US see extra latency (see Region and latency).
Integration surface	Libraries that you call from your code, frame by frame, inside your audio pipeline.	WebSocket protocol: a JSON client hello, then binary audio and JSON messages.
Platforms and languages	VIVA SDK: Python, Node.js, Go, Rust, C++ and C bindings, on Linux, Windows and macOS servers (Platforms). RTC SDK: servers, desktop, mobile (iOS, Android), desktop browsers, WebRTC and Electron (Platforms).	Any language or platform with a WebSocket client. Browsers pass the credential as a query parameter.
Network	The SDKs access the network only for license verification (see Security).	A connection to `streaming.krisp.ai` for the whole session.
Audio data	Processed on your premises (see Security).	Sent to Krisp over an encrypted (`wss://`) connection for processing.
Credentials	A commercial license from Krisp (see Licensing).	An API key from your Developers Portal dashboard. Your backend uses it to create short-lived session tokens, and clients send the session token in the `Authorization` header (see Authentication).
Products	VIVA SDK: Voice Isolation, Turn-Taking, VAD. RTC SDK: Accent Conversion, Noise Cancellation, Background Voice Cancellation, Voice Translation.	Speech-to-Text, Text-to-Speech, Voice Translation, Voice Isolation.

Voice Isolation, Turn-Taking and Voice Translation are available both ways. Each one's overview compares the two:

Voice Isolation: VIVA SDK or API endpoint
Turn-Taking: VIVA SDK or turn events in STT for Agents
Voice Translation: Voice Translation SDK or API endpoint

Choose an SDK if...

Audio must be processed on the device or on your own servers rather than sent to Krisp.
You need Noise Cancellation, Background Voice Cancellation, Accent Conversion, Turn-Taking or VAD. These are available in the SDKs.
You process audio frame by frame inside an existing audio path, for example WebRTC, Pipecat, LiveKit or a telephony stack.

Choose the API if...

You need Speech-to-Text or Text-to-Speech. These are available in the API.
You don't want to bundle native libraries or model files, or you work in a language the SDKs don't cover. Any WebSocket client works.
You prefer to run the models in Krisp's cloud instead of on your own CPU.

Next steps

Krisp API overview: authentication, the session protocol and the endpoint pages
VIVA SDK: Getting Started
RTC SDK: Getting Started