Sanas vs Krisp: Verified Facts about Accent Translation & Noise Cancellation Performance
Sanas thrives on competition, but only when it’s honest. Recently, Krisp has made claims that distort the facts about our technology, performance, and reach. Despite repeated requests for Krisp to remove the article, it remains online. We won’t let misinformation stand unchallenged.
The data speaks for itself: Sanas pioneered real-time speech understanding, and our results continue to set the global standard. The following comparison presents information on Krisp’s claims about Krisp capabilities and false and misleading claims about Sanas taken from a Krisp blog dated April 21, 2025, alongside the factss about Sanas as of October 2025, verified by live deployments and measurable outcomes.
Current Deployments
• Over 200 million desktop and mobile devices
• Over 200K contact center agents
• Over 1 trillion minutes of Krisp-processed voice
• Embedded into world-class services such as Vonage, RingCentral, Zoho, Aircall, Discord, others
• Over 30K agents
• 725k+ agents live throughout India, the Philippines, Latin America, Africa & Middle East
Accent Translation Robustness
Supported Accent Packs
• Indian English
• Filipino English
• Indian English
• Filipino English
• India
• Philippines
• Latin America
• Africa & Middle East
Across all live regions (Tier 1, Tier 2, and Tier 3 cities), over 99% of the English population is covered
Audio Latency
220ms
350ms-450ms
200ms
Modes of Operation
• Voice Preservation mode – fully preserves the user’s voice
• Voice Profiles mode – allows the user to choose a natural-sounding output voice
Voice Preservation mode – somewhat preserves the user’s voice
Fully preserves the speaker’s voice. We invented speaker identity preservation not as a feature, but as a principle
Accent Leakage
• Some leakage in Voice Preservation mode
• No leakage in Voice Profiles mode
Consistently observed leakage
No leakage and at the same time personalized to the user’s voice
Background Noise and Voice Cancellation Robustness
Highly robust, automatically included in the Accent Conversion models
Very limited
Noise and voice cancellation are included in Accent Translation
Agent and Customer-Side Noise Cancellation
Bi-directional, automatically included in the Accent Conversion models
Customer-side only
Bi-directional, automatically included in Accent Translation
Headset Robustness
Highly robust
Requires specific headsets
Highly robust and works well across all major headsets
Robustness across Users
Works consistently across all users
Requires testing three different versions for each user
Works consistently across all users
Wrong Pronunciations
Some
Noticeably more frequent
As measured using CEFR scores, Sanas significantly improves pronunciation scores
Preserves User’s Voice
Yes
Limited
Preserves the authenticity of each voice with unmatched fidelity
User Enrollment Needed
No
No
No
Dynamic Adaptation to New Speakers
Yes, within the same or different call, regardless of the gender
Unknown
Requires an output voice gender selection
Sanas preserves each person’s unique vocal identity; we don’t replace it with an avatar. While others rely on synthetic avatar voices that mimic or overwrite the speaker, Sanas keeps the user’s voice at the center. Our gender selection helps users sound closer to their authentic selves.
Voice Quality
16khz (wide-band, VOIP, industry-leading voice quality)
8kHz only
16khz (wide-band, VOIP, industry-leading voice quality)
Noise Cancellation Robustness
Voice Quality and Noise Cancellation
World’s best, based on objective and subjective tests
New entrant, tests show noise leakage and voice quality degradation
World’s best, based on objective and subjective tests (beat Krisp on all subjective metrics by a large margin)
Agent-Side Background Voice Cancellation
World’s best (see test measurements)
Other voices and background chatter leakage when in a typical loud call center
World’s best, scores compared to Krisp-BVC on our background voices challenge set:
• NISQA: +15.6%
• DNSMOS-sig: +13.3%
• DNSMOS-bak: +4.4%
• DNSMOS-ovr: +13.0%
• DNSMOS- p808: +0.1%
Agent-Side Noise Cancellation
World’s best (see test measurements)
• Adequate performance for low-volume noises (fan, for example)
• Noise leakage and voice degradation in contact center environments (other voices, loud chatter)
World’s best, scores compared to Krisp on our inbound challenge set covering a wide range of acoustics environments, noise type, and signal degradation:
• NISQA: +7.5%
• DNSMOS-sig: +8.6%
• DNSMOS-bak: +5.9%
• DNSMOS-ovr: +12.5%
• DNSMOS- p808: +7.5%
Customer-Side Noise Cancellation
Included
Optimized for inbound voice from mobile or landline. Pass-through of ringtones, dialtones, etc.
Not available
Included
Optimized for inbound voice from mobile or landline. Pass-through of ringtones, dialtones, etc
Acoustic Echo Cancellation
Included
Optimized for call center use cases
Not available
Not available
Voice Quality
• 8kHz (narrow-band, standard telephony, good voice quality)
• 16khz (wide-band, VOIP, industry-leading voice quality)
• 32kHz (full-band, best voice quality – near studio-grade)
8kHz only
8kHz, 16kHz, and 24kHz (wide-band, VOIP, industry-leading voice quality)
Application and Audio Drivers Robustness
CPU Utilization
• Supports range of CPUs typically in agent desktops
• Supports older, lower-end CPUs through smaller models
• Has auto-switching between models based on CPU load
• Single model uses 2x more than Krisp on i5-8th Gen CPU
• Error message in Sanas app with older CPUs
• Slightly higher CPU utilization for CPUs beyond i5 12th gen
• Uses half the CPU of Krisp for Accent Translation
• No errors
• Uses less memory and lower CPU load, making it possible to deploy on older laptops and thin-client setups.
• Leverages an efficient multithreading setup to remain stable and performant even under stressful compute conditions.
Audio Drivers
Highly reliable and tested for 7+ years
Users often need to restart the drivers to avoid breakdown of mic and speaker audio streams.
Reliable, industry-standard audio driver that runs with very low latency
Headset and Application Compatibility
Compatible and tested with most headsets and voice applications used in call centers
New entrant, minimal deployments and testing
Sanas is agnostic to application and supports most headsets used in the industry
Management and Deployment at Scale
Supported Platforms
Win, Mac, Linux, Chrome, VDI
Win
Win, Mac, Linux, IGEL, VDI
Installation Package
Single installation package including all accent packs and noise cancellation
• A separate package for different accent packs
• A separate package for noise cancellation
Single, multi-geo package that works across regions while supporting rapid deployment across one enterprise or a combination of enterprise and multiple BPOs.
SSO Authentication
• Available for agents, per the enterprise customers’ requirements
• SSO/SCIM for automated provisioning and deprovisioning, saving admins’ time
• Not available for users (agents)
• Only available for admins
• We support three options for all user types: SSO, Auto-Activation, and User Key.
• Auto-Activation is zero-touch (no integration) automated user provisioning
Remote Deployment and Settings for Admins
Highly Scalable
Very Limited
Highly Scalable
App Version Management and Auto-Update
Highly Scalable
Very Limited
Admins can review release details and push new versions with one click through the Sanas portal. Updates can be scheduled, bandwidth-throttled, and rolled out by user group. Profiles are retained, so agents move to the new version seamlessly, without IT overhead. Admin gets to control and publish a new version even when auto update is enabled.
Analytics for Accent Conversion, Noise Cancellation and Platform Usage
Available
Not available
Available through the administrator console
Enterprise-Grade Support
24/7
Application and IT infrastructure expertise during pilots and post-launch, including VDI
24/7
Limited
End-to-end enterprise support covering pilots, deployment, and post-launch operations. Includes dedicated engineering assistance for complex environments, VDI, and multi-region rollouts.
In-Depth Accent Translation Evaluation
Sanas has also conducted comprehensive objective evaluations on our latest Accent Translation model (4.0). These results — covering clarity, latency, and speaker identity preservation — are publicly available and independently verifiable. The findings demonstrate significant performance gains across every key metric. You can review the full results here: Meet Sanas Accent Translation 4.0: Clearer, Faster, More Human.
At Sanas, we don’t just correct the record, we keep moving it forward. Our mission has always been to make every voice understood, and that commitment continues to drive our innovation. With the recent launch of our real-time Language Translation mobile app, Sanas is extending its speech understanding technology beyond customer experience and into everyday global communication, continuing to lead the way in making connection truly universal.
