Improved Voice Settings for AI Agents and More with Cognigy.AI v4.96

The latest Cognigy.AI v4.96 release introduces new voice customization capabilities, allowing you to define and fine-tune voice settings at both Agent and Job levels.

Additionally, you can now leverage Anthropic’s newest Claude 3.7 Sonnet model for your AI Agents and benefit from improved call control with customizable duration limits upon transfer.

Tailored Voice Configuration for Every AI Agent and Job 

Cognigy continues to offer maximum flexibility in defining the voice persona of AI Agents. With native support for 1,000+ multilingual synthetic voices and speech recognition services from leading providers, including 11Labs, AWS, Deepgram, Google, Microsoft, Nuance, and Soniox, you can craft an AI voice experience that resonates with your audience.

Previously, voice configuration has been handled dynamically using the Session Config Node at any point within an Agent Flow. While effective, this requires, at times, additional configuration, e.g., when the same AI Agent is used for different Flows and tasks.

With Cognigy.AI v4.96, voice settings can now be seamlessly customized at the Agent and Job levels within the Agent Creation Wizard and AI Agent Node. This allows for improved manageability and more intuitive voice configuration, ensuring engaging and brand-aligned voice experiences for the intended use cases.

Native Support for Claude 3.7 Sonnet Model

In our quest to deliver cutting-edge LLM access, Cognigy.AI v4.96 also adds native support for Claude 3.7 Sonnet - the latest and most intelligent model from Anthropic capable of generating near-instant responses. According to Anthropic’s data, this model achieves industry-leading performance on the Tool-Agent-User (TAU) bench, an industry-standard framework for testing AI Agents on complex real-world tasks with user and tool interactions.

Another noticeable aspect of Claude 3.7 Sonnet is its improvement in balancing safety and reliability. It is allegedly better at grasping nuanced differences between malicious and benign inputs, lowering unwanted refusals by 45% compared to its predecessor.

Source: Anthropic

Improved Call Control with Customizable Duration Limits

The new Duration Limit setting in the Transfer Node and Voice Gateway Endpoint (when using the Dial option) allows you to specify a maximum duration value after a call has been transferred (e.g., to a human agent). This feature gives enterprises better control over call lengths, typically in scenarios where the end-user remains on the line but is no longer available. Once the set time is reached, the call disconnects automatically, regardless of whether the caller is still on the line.

This enhancement enables contact center managers to prevent excessive call times plaguing agent productivity, better allocate resources in time-sensitive customer interactions, and more efficiently manage costs, particularly in pay-per-minute services.

Other Improvements

Cognigy.AI 

• Added a limit of 100 Cognigy.AI API calls from the Code Node to prevent excessive API usage
• Added caption messages when sending attachments with Genesys Cloud Open Messaging

• Improved the tooltip text of the Failover Transcribe Enabled setting in the Call Failover section of the Voice Gateway Endpoint settings
• Added support for Elevenlabs's model Flash v2.5 for the Synthesizer (Text-to-Speech)
• Added the capability to transfer the call when an error occurs with the TTS vendor of the Synthesizer (Text-to-Speech) or Recognizer (Speech-to-Text)
• Added support for Elevenlabs's model Flash v2.5 to api-server
• Added the Voice over Internet Protocol (VoIP) carrier to the X-Requested-Carrier-Sid header for the sip:refer verb in outbound calls
• Added a maximum time-to-live (TTL) for sessions to ensure stale sessions are cleaned automatically. This enhances the overall system robustness and reduces potential memory leaks

Cognigy Webchat

• Added the Persistent menu to Webchat v3 and the corresponding settings to the Webchat v3 Endpoint settings
• Added the capability to handle typing indicators for Genesys Cloud Open Messaging and display when a human agent is typing in Webchat
• Added support for new Webchat v3 settings mode: Webchat v3 can be configured to render Markdown, collate streamed output messages, and display messages as progressive stream

For further information, check out our complete Release Notes here.