Krisp Voice AI SDK

What it does

Krisp Voice AI SDK is a set of real-time speech models that take in live audio (or text for speech synthesis) and return cleaner or transformed output. The VIVA SDK isolates the main speaker, detects voice activity and predicts turn endings and interruptions. The RTC SDK removes noise and background voices and converts accents. A separate cloud API streams speech-to-text, text-to-speech, voice translation and voice isolation over WebSockets.

Developers place the models in a voice pipeline: VIVA runs on their own servers in front of a speech-to-text or voice-activity step, the RTC SDK runs inside an app's audio path on the user's device, and the cloud API needs no library to ship. The models are small, run on CPU, work from audio alone without a transcription step, and each can be used independently. Bindings cover C, Python, Node.js, Go and Rust, plus LiveKit and Pipecat.

It suits engineers building voice agents, calling apps or contact-centre tools. Access is by SDK request, with an Early stage application and a separate API key for the cloud endpoints. Pricing for the SDKs is not published, and the self-serve translation API includes 60 free minutes. Several API products are labelled Beta.

What it can do

• Voice Isolation removes background noise and secondary voices before speech-to-text
• Turn Prediction and Interruption Prediction work directly from raw audio
• Voice Activity Detection tuned for on-server CPU use
• Noise and Background Voice Cancellation for inbound and outbound call audio
• Real-time Accent Conversion for human-to-human calls
• Cloud WebSocket API for speech-to-text, text-to-speech, voice translation and voice isolation

Works with

• LiveKit
• Pipecat

Example workflows

Clean caller audio for a voice agent

Input: live caller audio with background chatter. Steps: run Voice Isolation on your server ahead of VAD or speech-to-text, then add Turn Prediction and Interruption Prediction. Output: cleaner audio for transcription and better timing of when the agent speaks or stops.

Translate a live call

Input: streamed speech in one language. Steps: create a session token from your API key on your backend, open a WebSocket to the voice translation endpoint and stream audio. Output: translated speech and optional transcripts.

Before you choose

• VIVA models target human-to-AI agents, while RTC models target human-to-human calls; choose by use case
• SDK models run on your CPU infrastructure, whereas the cloud API sends audio to Krisp's servers
• Speech-to-text and text-to-speech API endpoints are marked Beta

Listed under

• Coding
• Call-centers
• Developers
• Noise-cancellation
• SDK
• Speech-translation
• Voice-agents

Get started

SDK access is by request form; API key from the dev dashboard

1. Request SDK access via the form, or apply for the Early stage plan
2. Create an API key in the developer dashboard if using the cloud API
3. Follow the VIVA SDK, Browser SDK or Krisp API quickstart in the docs

Developer pricing shows an application-based Early stage tier and a larger tier with volume pricing; no rates were confirmed. The Voice Translation API includes 60 free minutes, a limited allowance rather than a confirmed permanent free tier.

We couldn’t find published prices for Krisp Voice AI SDK. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• #1 Real-Time AI Voice SDK | Krisp ↗krisp.ai
• Krisp Developer Docs ↗sdk-docs.krisp.ai
• Krisp Pricing Plans ↗krisp.ai

Company website

krisp.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.