On-Demand AI-model Update

Introduction

On-demand AI model update lets you deploy specific AI models for Accent Translation, Speech Enhancement, and Noise Cancellation at the group level, directly from the Sanas Portal. Updates download in the background and apply automatically; agents stay on their calls without interruption.

AI model updates are independent of the app release cycle. You never need to push an app reinstall for a model change.

Feature overview

How it works

You can trigger an AI model update in two ways:

• Automatic update: Sanas recommends and deploys the optimal model when you change accent configurations (for the Sanas AT product).

• Manual update: You select a specific model version for any supported product.

User experience during an AI model update

What agents experience depends on whether the target accent changes.

Target accent changes (for example, US → UK)

AT pauses until the new AI model finishes downloading. During this time, Noise Cancellation remains active, and calls continue without interruption. Once the download completes, AT resumes automatically with the updated AI model.

Tip: Give your team a heads-up before switching target accents during active call hours.

No target accent change

When the target accent stays the same — for example, you're updating the model version, agents don't notice anything. AT and NC continue running on the current model while the new model downloads in the background. Once ready, the app switches to the new model on the next routine sync.

For admins: Configure AI models

Automatic AI model update

Use this method when you're configuring or changing accents for the Accent Translation product. Sanas recommends the optimal model automatically.

1. Sign in to the Portal and select the target group from the Group Tree. Verify that the group has access to the AT product.

2. Navigate to Settings > Configuration, then select Accent Translation under Products.

3. Configure the source and target accents based on your team's requirements.

4. Review the pre-selected recommended AI model, then select Save to deploy.

Note: Agents continue using Noise Cancellation while the new AT model downloads. AT becomes available again once the download completes.

Manual AI model update

Use this method when you want to select a specific model version for any supported product — Accent Translation, Speech Enhancement, or Noise Cancellation.

Common scenarios include:

• Piloting early-access models: Test a new AI model release on a small group before rolling it out broadly.

• Optimizing for hardware constraints: Deploy a lighter model for agents on lower-spec machines.

Steps:

1. (Optional) To access early-release models, enable Experimental Features under Settings > Configuration > Experimental Features.

2. Navigate to Settings > Configuration, then select the product you want to update (Accent Translation, Speech Enhancement, or Noise Cancellation).

3. Select your preferred AI model from the dropdown.
   

4. Select Save to deploy.

The model downloads to agent devices based on the group's configured Download Preferences.

Related articles

Support