Overview This guide will walk you through processing your first audio stream using the Sanas SDK.
This guide uses Sanas Cloud. For deployment options, see
Deployment .
Prerequisites
Make sure your environment meets the following system requirements before installing the SDK:
OS Version: Ubuntu 22.04 x86-64Python Version: 3.10 or higher Mac and Windows support coming soon.
To authenticate with Sanas Cloud, you’ll need an Account ID , Account Secret , and a Remote Endpoint URL . These credentials are provided in your welcome email after requesting access. If you don’t have credentials yet, request SDK access to get started. Step 1: Install the SDK
Download your SDK connector file (e.g., sanas_remote_sdk_<platform-arch>_<version>.tar.gz) included in the welcome email from Sanas.
Extract the archive:
tar -zxvf sanas_remote_sdk_ < platform-arc h > _ < versio n > .tar.gz
Change directory into the extracted folder:
cd < extracted-folder-nam e > # Replace with the actual folder name you see
The extracted folder name is usually similar to the tar.gz file name. Check the exact name after extraction.
Run the installer:
Step 2: Verify Installation Activate the SDK environment and test:
source sanas_remote_sdk/bin/activate python3 -c "import sanas_remote_sdk; print('SDK installed successfully!')"
This ensures your SDK is initialized correctly and ready for use.
Step 3: Select Your Model Choose the AI model that fits your use case:
Model Use Case Behavior Best For AGENTIC_VI_G_NCHuman ↔ Machine Full voice isolation Voice agents, IVR, STT AGENTIC_ST_NCHuman ↔ Machine Preserves distant speech Multi-speaker, panels, meetings VI_G_NC3.0Human ↔ Human Isolates intended speech Contact centers, conferencing, gaming
View full model comparison → Step 4: Process Your First Audio Stream Create a new Python file audio_stream_quickstart.py and copy the following code into the file.
Agentic Noise Cancellation (General)
Agentic Noise Cancellation (Standard)
Noise Cancellation (General)
import sanas_remote_sdk # Initialize SDK sdk = sanas_remote_sdk.CreateRemoteSDK() init_params = sanas_remote_sdk.InitParams() init_params.remoteEndpoint = "your_endpoint_url" # From credentials email init_params.accountId = "your_account_id" init_params.accountSecret = "your_account_secret" init_params.secureMedia = True result = sdk.Initialize(init_params) print ( f "SDK initialized: { 'SUCCESS' if result == sanas_remote_sdk.InitSDKResult. SUCCESS else 'FAILED' } " ) # State callback def state_callback ( state , reason ): if state == sanas_remote_sdk.ProcessorState. READY : print ( "Processor ready!" ) # Create an audio processor (full voice isolation) audio_params = sanas_remote_sdk.AudioParams() audio_params.modelName = "AGENTIC_VI_G_NC" audio_params.sampleRate = 16000 # Define source audio sampling rate processor, create_result = sdk.CreateAudioProcessor(audio_params, state_callback) if create_result == sanas_remote_sdk.CreateProcessorResult. SUCCESS : # Process audio chunks (dummy data) # 20 ms of audio at 16 kHz -> 20ms * 16kHz = 320 samples num_samples = 320 dummy_amplitude = 0.1 input_audio = [dummy_amplitude] * num_samples output_audio = processor.ProcessSamples(input_audio) print ( f "Processed { len (output_audio) } samples" ) # You can now send output_audio to your ASR system sdk.DestroyAudioProcessor(processor) sdk.Shutdown()
import sanas_remote_sdk # Initialize SDK sdk = sanas_remote_sdk.CreateRemoteSDK() init_params = sanas_remote_sdk.InitParams() init_params.remoteEndpoint = "your_endpoint_url" # From credentials email init_params.accountId = "your_account_id" init_params.accountSecret = "your_account_secret" init_params.secureMedia = True result = sdk.Initialize(init_params) print ( f "SDK initialized: { 'SUCCESS' if result == sanas_remote_sdk.InitSDKResult. SUCCESS else 'FAILED' } " ) # State callback def state_callback ( state , reason ): if state == sanas_remote_sdk.ProcessorState. READY : print ( "Processor ready!" ) # Create an audio processor (preserves background speech) audio_params = sanas_remote_sdk.AudioParams() audio_params.modelName = "AGENTIC_ST_NC" audio_params.sampleRate = 16000 # Define source audio sampling rate processor, create_result = sdk.CreateAudioProcessor(audio_params, state_callback) if create_result == sanas_remote_sdk.CreateProcessorResult. SUCCESS : # Process audio chunks (dummy data) # 20 ms of audio at 16 kHz -> 20ms * 16kHz = 320 samples num_samples = 320 dummy_amplitude = 0.1 input_audio = [dummy_amplitude] * num_samples output_audio = processor.ProcessSamples(input_audio) print ( f "Processed { len (output_audio) } samples" ) # You can now send output_audio to your ASR system sdk.DestroyAudioProcessor(processor) sdk.Shutdown()
import sanas_remote_sdk # Initialize SDK sdk = sanas_remote_sdk.CreateRemoteSDK() init_params = sanas_remote_sdk.InitParams() init_params.remoteEndpoint = "your_endpoint_url" # From credentials email init_params.accountId = "your_account_id" init_params.accountSecret = "your_account_secret" init_params.secureMedia = True result = sdk.Initialize(init_params) print ( f "SDK initialized: { 'SUCCESS' if result == sanas_remote_sdk.InitSDKResult. SUCCESS else 'FAILED' } " ) # State callback def state_callback ( state , reason ): if state == sanas_remote_sdk.ProcessorState. READY : print ( "Processor ready!" ) # Create an audio processor audio_params = sanas_remote_sdk.AudioParams() audio_params.modelName = "VI_G_NC3.0" audio_params.sampleRate = 16000 # Or up to 24kHz processor, create_result = sdk.CreateAudioProcessor(audio_params, state_callback) if create_result == sanas_remote_sdk.CreateProcessorResult. SUCCESS : # Process audio chunks (dummy data) # 20 ms of audio at 16 kHz -> 20ms * 16kHz = 320 samples num_samples = 320 dummy_amplitude = 0.1 input_audio = [dummy_amplitude] * num_samples output_audio = processor.ProcessSamples(input_audio) print ( f "Processed { len (output_audio) } samples" ) # You can now send output_audio to your ASR system sdk.DestroyAudioProcessor(processor) sdk.Shutdown()
Step 5: Run the Code Run the code and the console will print out the SDK initialization and the processed samples.
python audio_stream_quickstart.py
Next Steps Now that you’ve successfully processed a single audio stream with the Sanas SDK, you can move on to handling multiple concurrent streams.
Processing Multiple Streams Scale to multiple concurrent audio streams with a shared SDK instance.
References
RemoteSDK SDK lifecycle methods including Initialize, CreateAudioProcessor, and Shutdown.
AudioProcessor ProcessSamples and GetState methods for audio processing.
AudioParams Model selection and sample rate configuration.
InitParams SDK connection configuration including endpoint and credentials.
Need Help?
Support Portal Raise a support ticket for urgent issues
