Skip to main content
AI-powered speech processing for voice agents and real-time communication

​
What is Sanas?

Sanas provides speech solutions powered by our AI models that enhance voice communication quality and improve speech recognition accuracy in real-world noisy environments.

Quick Start

Get started in 5 minutesInstall the SDK and enhance your first audio stream

Explore Solutions

See what you can buildBrowse use cases and audio samples

​
What You Can Build

Improved Audio Pipeline for AI Voice Agents

Improve ASR accuracy by 5-30% in noisy environmentsUse cases: Voice assistants, IVR systems, chatbots→ Start building

Noise Cancellation

Remove background noise from human conversationsUse cases: Contact centers, conferencing, gaming→ Start building

​
Choose Your Integration Path

You can start building with Sanas through the following integration paths:
Sanas SDK - Cloud-Hosted Solution
  • Real-time streaming
  • Low latency
  • Sanas Cloud or self-hosted options
Best for:
  • Real-time applications
  • Voice pipelines
  • Contact centers

Get Started

Install and run your first example

API Reference

Complete SDK documentation
When to use SDK:
  • You need real-time streaming with low latency
  • You want full control over the audio processing pipeline
  • You need offline processing mode
Here are some popular use cases for our speech solutions:
Challenge: Your voice agent misinterprets user input in noisy environments (cafes, streets, offices).Solution: Use our AI Voice Agent model to reduce Word Error Rate (WER) by 5-30%.Quick Start:
# 30-second example
from sanas_sdk import create_sdk

sdk = create_sdk(endpoint="...", account_id="...", secret="...")
processor = sdk.create_processor(model="AGENTIC_NC")

clean_audio = processor.process(noisy_audio)
# Pass to your ASR (Google, Deepgram, etc.)
Full Tutorial →
Challenge: Contact center calls are often accompanied by background chatter, keyboard noise, and office sounds.Solution: Use the Noise Cancellation model to isolate the speaker’s voice, delivering a crystal-clear calling experience. Quick Start:
from sanas_sdk import create_sdk

sdk = create_sdk(endpoint="...", account_id="...", secret="...")
processor = sdk.create_processor(model="VI-G", confidence=1.0)

clean_audio = processor.process(noisy_call_audio)
Full Tutorial →
Challenge: Handle 100+ concurrent calls in a contact center.Solution: Sanas SDK supports concurrent streams with consistent, robust performance. Multi-Stream Tutorial →
Challenge: Regulatory requirements prevent sending audio to external servers.Solution: Deploy the self-hosted Sanas SDK on your infrastructure to maintain complete control and meet your compliance needs. Contact Support →

​
Support & Resources

Contact Support

Get help from our team

View Changelog

Latest updates & fixes