Spectre TTS Engine

Converse uses a high-quality text-to-speech service integrated natively into the platform. It is designed for conversational AI with ultra-low latency and natural-sounding output across 20+ languages.

Voice Library

A large library of realistic voices is available, spanning multiple languages, accents, and tones. Voices are organized by language and available in the voice picker when configuring an agent.

• Language support: English (US/UK/India), Hindi, Tamil, Telugu, Malayalam, Kannada, Marathi, Spanish, French, German, Arabic, and more.
• Voice styles: Professional, conversational, warm, energetic — choose the one that fits your brand.
• Preview: Use the Playground → TTS tab to preview any voice before assigning it to an agent.

Latency

The TTS service is optimized for real-time voice conversations. It uses a streaming WebSocket connection — audio starts playing to the caller as soon as the first sentence is synthesized, without waiting for the full response. This dramatically reduces perceived latency.

Language detection

When you set an agent's language, the TTS will speak in that language regardless of the text it receives. If your system prompt instructs the agent to respond in Hindi, the Hindi voice will be used automatically. Configure the voice language to match the expected caller language.

Voice Cloning

Custom voice cloning is available on paid plans. Provide 15–30 seconds of clean audio from a speaker, and a cloned voice is generated within minutes. Cloned voices can be used exactly like library voices — select them from the voice picker in any agent.

Testing TTS

Go to Playground → TTS tab to enter any text and hear it spoken in your chosen voice. You can adjust speed and test different voices side by side. The Playground → Flow tab also plays TTS audio for every agent response when you test a flow.