General Assistant Settings

​

Quick Start

How to set up your first Famulor assistant:

1. Choose call direction: Incoming (receive calls) or outgoing (make calls).
2. Assign assistant name: Internal name such as “Support Assistant” or “Sales Qualifier.”
3. Configure phone numbers: Assign platform numbers, SIP numbers, or just a verified caller ID number.
4. Select voice & language: Use built-in voices or clone your own voice.
5. Adjust advanced settings: Fine-tune model, timing, audio parameters, and protection mechanisms.

​

Call Direction & Basic Setup

​

Assistant Type

Specify whether your assistant handles incoming or outgoing calls. This choice determines which number and routing options are visible in Famulor.

• Incoming (receive calls): Your assistant answers calls from customers, e.g., as a digital switchboard or support hotline.
• Outgoing (make calls): Your assistant actively calls leads or existing customers, e.g., for qualification, follow-ups, or appointment confirmations.

​

Assistant Name

Choose a clear, descriptive name that reflects the purpose and use case, for example:

• “Outbound Lead Qualification B2B”
• “Customer Support Level 1”
• “Appointment Booking Dr. Müller’s Practice”

A consistent naming scheme helps your team quickly identify assistants within the Famulor interface.

​

Phone Number Configuration

Your assistant needs at least one assigned phone number to be reachable in the telephone network or to make calls.

​

For Outgoing Assistants

You can use the following options:

• Platform numbers: Numbers booked directly through Famulor, ideal for quick setup without your own phone system.
• SIP numbers: Integration of your existing VoIP or PBX infrastructure via SIP trunks.
• Caller ID only: An existing company number is verified and used solely as the displayed caller ID on outgoing calls.

​

For Incoming Assistants

Typically use:

• Platform numbers: You route customers directly to a number provided by Famulor.
• SIP numbers: You connect your existing phone system with Famulor and forward calls to the assistant.

Important: Caller ID–only numbers cannot receive incoming calls. They are used only as visible caller ID on outgoing calls.

​

Cost Overview

• Platform numbers: Monthly base fee per number plus usage-based per-minute prices.
• SIP integration: Usually no monthly fee from Famulor; only per-minute charges for AI processing apply.
• Caller ID: No monthly fees; per-minute charges apply for outgoing calls.

For details on prices, regions, and types, refer to the “Phone Numbers” and “Costs & Pricing” sections in the documentation.

​

Engine Type (Speech Processing Mode)

Determine how your Famulor assistant processes speech and generates responses. The mode choice affects latency, naturalness, and configuration options.

​

Pipeline Mode

Classic chain: speech recognition → language model → speech output. Suitable for:

• Complex reasoning and long responses
• Precise control over voices, transcriber, and models
• Scenarios with many tools and integrations

​

Speech-to-Speech Mode

Direct speech-to-speech processing without visible intermediate text. Suitable for:

• Very natural conversations
• Fastest turn-taking experience
• Simple use cases focused on conversational flow

​

Dualplex Mode (Beta)

Combines fast multimodal processing with high-quality voices. Suitable for:

• Standard use cases prioritizing performance and quality
• Recommended default for many Famulor assistants

​

Language Configuration

​

Primary Language

Define the primary language your assistant speaks and understands. This setting affects:

• Accuracy of speech recognition
• Available voices and accents
• Default filler phrases
• Selection of the appropriate model

​

Secondary Languages

Optionally, your assistant can understand and speak additional languages. Typical use cases:

• Multilingual customer support
• International companies operating in multiple countries
• Conversations switching between languages

The assistant usually detects automatically which language the other person speaks and responds accordingly.

​

TTS Provider & Voice Selection

​

TTS Provider

In Pipeline and Dualplex modes, you can select the text-to-speech provider. Typical options include:

• High-quality premium voices with many accents
• Fast, low-latency voices for reactive conversations

​

Voice Options

There are three ways to find the ideal assistant voice:

1. Use existing voices
   • Pre-made professional voices
   • Different accents and speaking styles
   • Various gender and tonal variants
2. Clone your own voice
   • Upload voice recordings of a speaker.
   • Minimum recording length and quality requirements apply (one person, minimal background noise).
   • After processing, you can select the voice directly in the assistant.
3. Tune brand voice
   • Choose and configure a voice that matches your brand (formal, casual, friendly, businesslike).
   • A consistent voice across multiple assistants builds brand recognition.

​

Time Zone Settings

​

Time Zone

Your assistant’s time zone controls:

• Time references in conversations (“today,” “tomorrow,” times)
• Appointment scheduling logic
• Timestamps in call logs and reports

Recommendation: Choose the time zone of your main location or the majority of your customers for consistent time references.

​

Audio Optimization

​

Background Noise

Optionally, you can mix a subtle ambient noise under the voice, such as light office ambiance.

• Advantage: Feels more natural and masks minimal processing delays.
• Note: Too loud background noise can impair speech recognition.

​

Filler Audio

Filler audios are short confirmation sounds and phrases like “mhm,” “okay,” “understood” played during processing. Benefits:

• Prevent unnatural pauses in speech
• Keep callers engaged and reduce drop-offs
• Make the assistant sound more attentive

Famulor provides appropriate default filler phrases for each supported language. You can customize these by category (positive, neutral, questioning, etc.) to match your brand voice.

​

Advanced Settings

​

LLM Model

Choose the language model appropriate for your mode and use case:

• Models with low latency for very fast conversations.
• Models with strong reasoning for complex logic and long answers.

Practical tip:

• When response speed is crucial (e.g., support hotline or outbound campaigns), prioritize fast models.
• For complex decisions (e.g., multi-stage qualification, extensive consulting), use models with better reasoning and compensate latency with filler audio.

​

LLM Temperature

Range: 0.0 – 1.0  |  Default: 0.1 Temperature controls how “creative” and varied responses are:

• Low (0.0–0.3): Stable, repeatable, ideal for business and functional calls.
• High (0.7–1.0): Freer, more creative, e.g., for casual conversations.

​

Duration and Timing Settings

Control call duration, wait times, and silence to balance costs and user experience:

• Re-engagement interval: How many seconds of silence before the assistant prompts again.
• Re-engagement prompt: Optional phrasing for follow-up after silence (e.g., polite “Are you still there?”).
• Maximum call duration: Upper limit for total call length to control costs.
• Maximum silence duration: Time after which the call ends if no response is received.
• Ring duration: How long the phone rings before the call is marked unanswered (helps to avoid voicemail).

​

Protection and Quality Settings

These options help you manage call quality and costs:

• Noise suppression: Filters callers’ background noise. Disable if voices sound “cut off.”
• End on voicemail: Automatically ends the call when an answering machine is detected to avoid unnecessary costs.
• Voicemail message: Optionally leave a brief message (e.g., name, company, callback request).
• Record calls: Records calls for quality assurance and analysis – be mindful of local regulations and consents.
• Maximum initial silence: Ends the call if no one speaks at the beginning (e.g., when only voicemail picks up).

​

Synthesizer Settings

In Pipeline and Dualplex modes, you can fine-tune your assistant’s voice parameters.

​

Voice Modulation

• Emotional modulation: When enabled, adjusts tone and emphasis based on context and conversation.
• Stability: Controls how consistent the voice sounds – low values are livelier, high values calmer and more even.
• Voice similarity: Relevant for cloned voices – controls how closely output matches the original voice.
• Speech rate: Choose whether the assistant speaks slowly, normally, or quickly based on your target audience.

​

Transcriber Settings

In Pipeline mode you select the speech recognition service and endpointing behavior.

​

Transcriber Selection

Typically, multiple providers with different strengths are available:

• Providers offering maximum accuracy for critical professional calls.
• Providers with very low latency for reactive dialogues.

Test with your language, accents, and typical noise environments to find the most reliable service in Famulor.

​

Endpointing & Interruptions

Endpointing controls when the assistant stops listening and begins speaking:

• Endpoint sensitivity: How long to wait after the last word before the assistant responds.
• Interruption sensitivity: How easily the assistant stops if the caller interrupts.
• Minimum word count for interruption: Prevents random background noises from interrupting the conversation.
• Turn detection threshold (for multimodal modes): Controls how quickly the assistant detects the other party has finished speaking.

---