Configure the voice

Voice provider

The voice provider determines the technology used to generate the assistant's voice. You can choose from:

• ElevenLabs — Default provider. Offers the best Spanish voice quality with natural, expressive voices. You can select a specific voice ID and generation model (default: eleven_turbo_v2_5).
• OpenAI — Voices generated by OpenAI. Good quality, though less natural than ElevenLabs for Spanish.

For each provider, you need to configure:

• Voice ID — The unique identifier of the selected voice within the provider.
• Model — The provider's voice generation model. Each provider has its own available models.

Voice settings

Within the voice configuration, you can adjust two key parameters that control the generated voice's behavior:

• Stability — Value between 0 and 1. A higher value produces a more consistent, stable voice. A lower value allows more variation and expressiveness, but can be less predictable. Recommended: 0.5-0.7 for a balance between naturalness and consistency.
• Similarity boost — Value between 0 and 1. A higher value makes the voice more similar to the original reference voice. A lower value allows more deviation. Recommended: 0.75 to maintain fidelity to the original timbre.

Transcriber (speech-to-text)

The transcriber is the service that converts the patient's speech into text so the assistant can process it. Configure the following parameters:

• Provider — The transcription service. Default: Deepgram, which offers excellent accuracy and low latency.
• Model — The transcription model. Default: Nova-3, Deepgram's most advanced model for Spanish.
• Language — The transcription language. Set to Spanish by default. Not recommended to change unless your clinic serves in another language.

AI model

The AI model is the assistant's «brain» that processes the conversation and generates responses. Configure:

• Provider — The AI model provider. Default: OpenAI.
• Model name — The specific model to use. Default: gpt-4o-mini, which offers an excellent balance between response quality and speed.
• Temperature — Controls response creativity. Value between 0 and 1. A low temperature (0.2-0.4) produces more predictable, conservative responses. A high temperature (0.7-1.0) produces more varied, creative responses. Default: 0.5.

First message

The first message is what the assistant says when answering the call. You can fully customize it using free text with placeholders:

• {clinic_name} — Automatically replaced with your clinic's name.
• {assistant_name} — Automatically replaced with the assistant name configured in general settings.

Example first message:

«Hello, thank you for calling {clinic_name}. My name is {assistant_name}, how can I help you today? This call may be recorded to improve our service.»

Maximum call duration

Defines the maximum time a call can last. If this limit is reached, the call ends automatically. The default value is <strong>600 seconds (10 minutes)</strong>.

Adjust this value based on your clinic's needs:

• Clinics with quick appointments — 300-480 seconds (5-8 minutes) may be sufficient for most booking calls.
• Clinics with complex services — 600-900 seconds (10-15 minutes) allow longer conversations with detailed queries about services and availability.