Krisp Voice Isolation
krisp.ai
- Free trial
- Windows
- macOS
- Linux
- Web
- iPhone and iPad
- Android
- API
What Krisp Voice Isolation is
Krisp Voice Isolation is a real-time voice model in Krisp VIVA, a voice infrastructure offering for voice AI agents. It separates the primary speaker from other voices, background chatter, and room echo before speech reaches a speech-to-text engine. Krisp says Voice Isolation 2.5 runs on CPU with 15 ms latency and reduced word errors by 46% in its benchmark. The VIVA SDK also includes Turn Prediction, Interruption Prediction, Voice Activity Detection, and detectors for accent, gender, and synthetic speech. It runs server-side on-premises or in a customer's cloud, and is designed to sit before the speech-to-text engine. Listed integrations include LiveKit, Pipecat, and raw WebRTC pipelines; Krisp says the SDK works with any STT engine. Python, Node.js, Go, and iOS SDKs are listed, with PCM S16LE audio at 16 kHz mono in and out. The broader AI Voice SDK library supports Windows, Mac, Linux, web, iOS, and Android. Teams can describe their use case to request a trial; no duration is stated. Pricing is on request.
Who it suits
Krisp VIVA is aimed at teams building voice AI agents, including startups and early-stage builders experimenting with Voice AI. Its server-side deployment and Enterprise support options may suit teams with deployment or service-level requirements.
Good to know, and what to watch
- Separates a speaker from competing voices and room echo
- SDK includes turn, interruption, and activity models
- Runs server-side on-premises or in a customer's cloud
- Works with any STT engine
- Enterprise plan includes dedicated engineering support
- Pricing is available on request
- Trial access requires describing the use case
- Listed audio format is PCM S16LE, 16 kHz mono
Plans
Every plan and priceEarly stage
Enterprise
Pricing on request
Getting started
- Open the Krisp VIVA developer website
- Describe your use case to request SDK and trial access
- Choose Early stage or Enterprise based on required capabilities
- Select a listed SDK or supported integration
- Send and receive PCM S16LE 16 kHz mono audio
- Deploy VIVA server-side on premises or in your cloud
Limits to check
The Early stage plan lists VIVA and RTC models but does not list customization, batch processing, dedicated engineering support, or custom SLAs. Enterprise adds those capabilities, while pricing for both plans is available on request. The trial has no stated duration.
The Club's review
Overview
Krisp Voice Isolation is a real-time audio model that separates a main speaker from competing voices and room echo before speech-to-text. It is best suited to teams building voice AI agents that need cleaner audio and related speech signals; its server-side deployment and support for different STT engines are appealing, provided its audio requirements fit the pipeline.
Voice Isolation 2.5 is designed to run on CPU with 15 ms latency, and the VIVA SDK adds models for turn-taking, interruptions, voice activity, and signal detection. Krisp offers a trial setup for teams that describe their use case, but the trial has no stated duration.
Key features
Isolation is intended to clean audio before transcription, separating the primary speaker from other voices, background chatter, and room echo. Krisp says Voice Isolation 2.5 reduced word errors by 46% in its benchmark. The combination of low claimed latency and CPU operation is relevant for responsive voice agents, though benchmark results do not guarantee the same outcome in every deployment.
VIVA includes Turn Prediction, Interruption Prediction, Voice Activity Detection, and signal detectors for accent, gender, and synthetic speech. Those additions make the SDK more useful to teams shaping agent behavior than an isolation model alone. It is designed to sit before an STT engine, and Krisp says it works with any STT engine.
Teams can run VIVA server-side on-premises or in their own cloud. Krisp lists LiveKit, Pipecat, and raw WebRTC pipelines as supported integration paths. The SDKs listed for developers are Python, Node.js, Go, and iOS; the AI Voice SDK library is also available for Windows, Mac, Linux, Web (JS/WASM), iOS, and Android. Audio must be PCM S16LE, 16 kHz mono in and out, a concrete integration constraint that should be checked before adoption.
Pricing
Krisp uses paid, custom pricing and offers two developer plans. The Early stage plan includes VIVA and RTC models and is aimed at startups, early-stage builders, and teams experimenting with Voice AI. Teams apply for access rather than selecting a published price, so it suits projects willing to discuss their use case with Krisp before deployment.
Enterprise pricing is custom and volume-based. It adds higher-quality models, server-side batch processing, model customization, dedicated engineering support, and custom SLAs. This is the better fit for organizations that need tailored models, batch workflows, or contractual support; smaller teams that need only real-time isolation may not need those additions. Trial access is arranged after teams describe their use case, with no stated duration.
Platforms
Krisp lists Windows, Mac, Linux, Web, iOS, and Android support for its AI Voice SDK library. The VIVA page separately lists Python, Node.js, Go, and iOS SDKs. This range gives developers several routes into an integration, while the specified mono PCM format remains an important pipeline requirement.
Who it's for
VIVA is aimed at teams building voice AI agents, particularly those that want to clean speech before transcription and use additional signals for turn-taking, interruptions, or voice activity. On-premises and customer-cloud deployment provide options for teams choosing where server-side processing runs. It is less suited to individuals seeking a simple desktop noise-cancellation app or buyers who need a fixed, self-serve price before engaging with a vendor.
Krisp Technologies, Inc. identifies its headquarters as Berkeley, California. For security practices, compliance standards, policies, controls, and documentation, it directs customers to its Trust Center.
Pros and cons
- Pros: Isolation is designed to remove competing voices and room echo before transcription, addressing a specific source of poor voice-agent input.
- Pros: The SDK combines isolation with prediction, activity detection, and signal detectors, giving voice-agent teams related models in one offering.
- Pros: On-premises or customer-cloud deployment and support for LiveKit, Pipecat, raw WebRTC, and any STT engine allow several integration approaches.
- Cons: Custom pricing and an application-based Early stage plan make it harder to assess cost without contacting Krisp.
- Cons: PCM S16LE at 16 kHz mono is a firm audio-format requirement that may call for pipeline adjustments.
Alternatives
For a broader set of choices, browse AI Voice Isolators. Consider AudioShake if you want a free plan with 200 starting credits, MP3 export, and support for files up to 10 minutes. LALAL.AI is another option if a free tier with 10 minutes in a relaxed queue and 200 MB uploads suits your needs.
AIVoiceSeparator may suit users seeking a free plan with three songs per month and full Studio-quality output. Choose clir Focus if you want an iOS app with three free sessions before a subscription. Altered Real-Time Pro offers a free tier with 20 minutes of daily processing and speech-enhancement isolation models for macOS and Windows.
Demixr is a web alternative with a free allowance of 10 minutes of uploaded audio per month, subject to a 10 MB upload limit. Vocly offers an Android free tier with separation, mixer, karaoke mode, and a stem library. For iOS hearing assistance, clir Hearing has a monthly plan at $12.99 a month and a yearly plan at $59.99 a year.
Verdict
Krisp Voice Isolation is a strong fit for teams building voice agents that need to clean audio before transcription and want related voice models, flexible server-side deployment, and several integration paths. Its main reasons to look elsewhere are custom pricing, an audio-format requirement that may not match an existing pipeline, and trial access that requires a use-case discussion. Choose it when those integration and model capabilities matter more than self-serve pricing; otherwise, compare alternatives with a free tier or clearer entry terms.
The verdict
Krisp Voice Isolation is designed to clean audio before speech-to-text, with additional voice models and support for several integration paths. Teams should confirm trial access and pricing for their use case.
Questions people ask
Is Krisp Voice Isolation free?
No. Krisp lists a paid pricing model, with pricing available on request. Trial access is available after a team describes its use case.
Which platforms are supported?
The AI Voice SDK library is listed for Windows, Mac, Linux, web, iOS, and Android. VIVA also lists Python, Node.js, Go, and iOS SDKs.
Can it work with my speech-to-text engine?
Krisp says the SDK works with any STT engine and is designed to run before speech-to-text processing.
Where can VIVA run?
It runs server-side on premises or in a customer's cloud.
What does Enterprise add?
Enterprise lists higher-quality models, server-side batch processing, model customization, dedicated engineering support, custom SLAs, and volume pricing.
What the maker publishes
- Additional models
- The VIVA SDK includes Turn Prediction, Interruption Prediction, Voice Activity Detection, and Signal Detectors for accent, gender, and synthetic speech. source
- Company
- Krisp's site identifies the company as Krisp Technologies, Inc. and lists its address in Berkeley, California. source
- Deployment
- VIVA runs server-side on-premises or in a customer's cloud and is designed to sit before the speech-to-text engine. source
- Founded
- 2017 source
- Headquarters
- Berkeley, California, United States source
- Integrations
- Krisp lists LiveKit, Pipecat, and raw WebRTC pipelines as supported integrations, and says the SDK works with any STT engine. source
- Performance
- Krisp says Voice Isolation 2.5 delivers 15 ms latency, runs on CPU, and reduced word errors by 46% in its benchmark. source
- Purpose
- Krisp VIVA provides real-time voice models that clean, understand, and transform audio for voice AI agents. source
- SDKs and audio format
- The VIVA page lists Python, Node.js, Go, and iOS SDKs and specifies PCM S16LE, 16 kHz mono audio in and out. source
- Security
- Krisp directs customers to its Trust Center for details about security practices, compliance standards, policies, controls, and security documentation. source
- Support
- The Enterprise developer plan includes dedicated engineering support and custom SLAs. source
- Supported platforms
- Krisp says its AI Voice SDK library is available for Windows, Mac, Linux, Web (JS/WASM), iOS, and Android. source
- Trial access
- Krisp says teams can describe their use case to be set up with the SDK and a trial; the page does not state a trial duration. source
- Voice isolation
- Voice Isolation separates the primary speaker from competing voices, background chatter, and room echo before speech-to-text processing. source
- Who it is for
- Krisp describes VIVA as voice infrastructure for voice AI agents and says its Early stage plan is aimed at startups, early-stage builders, and teams experimenting with Voice AI. source
Checked against the maker's pages on 8 Oct 2026.