What models are and why they matter

When a customer calls, the assistant constantly does two things: understands what is being said and replies in a natural voice. Behind these two capabilities are two distinct models, and yourang.ai lets you choose which to use.

Two layers, two decisions

• Voice model. It's the voice the assistant uses to speak to the customer. It determines timbre, accent, rhythm, and naturalness of speech. Think of the voice model as your business's voice actor.
• AI model. It's the brain that understands requests, decides how to respond, and when to trigger actions (transfer, book, collect data). It determines how accurate, fast, and capable the assistant is at handling complex conversations.

The two choices are independent: you can combine a premium voice with a standard AI model, or vice versa. The right combination depends on your budget, your customer base, and the complexity of the requests you receive.

What differs between models

• Perceived quality. More advanced models sound more human, understand regional accents, and handle interruptions and overlaps naturally. On a reception call, they make the difference between a smooth experience and a slightly stiff one.
• Response speed. Some models reply in a few hundred milliseconds, others take a moment longer. For quick conversations (a customer asking about hours, another looking for a menu) latency is noticeable.
• Supported languages. Not all models speak every language at the same level. If your customer base is international, quality in English, French, German, or Spanish is an important criterion.
• Cost per minute. More capable models have a higher per-minute cost. It's a balance: for simple, high-volume requests, a standard model can be more than enough.

Models evolve over time

yourang.ai constantly updates its catalogue of voice and AI models as providers release new versions that are more capable, more natural, or faster. New additions are announced on the What's New page of the documentation and in the control panel. You can always switch to a new model from the agent configuration without recreating anything.