Krisp is a software company specializing in artificial intelligence-based noise cancellation and voice enhancement for communication platforms. Founded in 2017, the company provides a desktop application and SDK that removes background noise, echoes, and reverberation from both incoming and outgoing audio in real-time, improving call quality across various conferencing tools. Its core technology leverages deep learning models to separate human speech from ambient sounds without requiring specialized hardware.
Krisp's solutions are used by individuals, remote teams, and enterprises to enhance productivity in virtual meetings, webinars, and customer support interactions. The company has integrated its technology into numerous communication platforms and offers both consumer and business tiers, with a focus on low-latency processing and privacy preservation.
Technology
Krisp's noise cancellation is built on Deep learning and Neural network architectures, specifically Residual Network (ResNet) and U-Net models that process audio spectrograms. The system analyzes incoming audio frames, identifies non-speech components such as keyboard clicks, dog barks, or traffic, and suppresses them while preserving the speaker's voice. This is achieved through a Sequence-to-Sequence (Seq2Seq) approach that maps noisy audio to clean speech, trained on large datasets of paired noisy and clean recordings.
The company employs Data Augmentation techniques to simulate diverse acoustic environments, improving model robustness. Krisp also uses Batch Normalization and Dropout to stabilize training and prevent overfitting. The processing runs locally on the user's device, ensuring low latency (typically under 10 milliseconds) and no audio data leaves the device, addressing privacy concerns.
Krisp's models are optimized for real-time inference, leveraging Model Pruning and efficient computation to run on standard CPUs without requiring dedicated accelerators. The company continuously updates its algorithms to handle new noise types and improve voice clarity, including features like echo cancellation and voice enhancement for speakers with lower audio quality.
History
Krisp was founded in 2017 by Davit Baghdasaryan and Artavazd Minasyan, who previously worked on speech recognition and audio processing. The company initially launched as a mobile app for call recording and transcription but pivoted to noise cancellation after recognizing the growing demand for remote communication tools. In 2019, Krisp released its first desktop application for Windows and macOS, gaining rapid adoption among remote workers.
In 2020, Krisp raised a $5 million seed round led by OpenAI's startup fund, followed by a $35 million Series A in 2021. The funding supported expansion of its engineering team and development of enterprise features. By 2022, Krisp had integrated its SDK into major platforms like zoom (not in list, but implied), microsoft-teams (not in list), and google-meet (not in list), and reported over 100 million noise-free minutes processed monthly.
Products and Integration
Krisp offers a standalone desktop app for Windows, macOS, and Linux that works with any communication application, including video conferencing, VoIP, and online gaming. The app provides real-time noise cancellation for both microphone input and speaker output, with adjustable intensity levels. It also includes voice enhancement features that improve clarity for quiet speakers.
For developers, Krisp provides a software development kit (SDK) that can be embedded into third-party applications. The SDK supports multiple programming languages and platforms, enabling integration into Amazon Web Services and Microsoft Azure cloud-based services. Krisp has partnered with hardware manufacturers to pre-install its technology on laptops and headsets, and offers a business tier with centralized management and analytics.
Market and Impact
The rise of remote work and virtual meetings has increased demand for audio quality solutions. Krisp competes with other AI-based noise cancellation providers, but differentiates through its on-device processing and broad compatibility. The company's technology has been adopted by customer service centers, educational institutions, and healthcare providers to improve communication clarity.
Krisp's impact extends to accessibility, as its voice enhancement helps individuals with speech impairments be better understood. The company also emphasizes privacy, as all processing occurs locally, contrasting with cloud-based alternatives. As of 2024, Krisp continues to expand its product line, including features for real-time transcription and meeting summaries, leveraging Generative AI and Large language model technologies.