Setup and Implementation
Implementation starts with generating an API key from your account dashboard, then deciding your model tradeoff (Multilingual v2 for quality, Turbo/Flash for latency) and your voice source (Voice Library or a clone) before writing any integration code -- these choices affect the API calls you'll make, so it's worth deciding them before building.
Practical Details
For a Conversational AI agent specifically, setup also includes configuring the agent's function-calling tools and, if needed, telephony connection -- a meaningfully bigger setup task than a simple TTS integration, which is often a single API call per piece of text.
Actionable Considerations
- Decide your model (quality vs. latency) and voice source before writing integration code, since switching later means re-testing output quality.
- Start with the API playground or a small test script to validate output before building a full production pipeline.
- For Conversational AI agents, budget extra setup time for function-calling and telephony configuration beyond basic TTS integration.