Setup of an AI Voice Bot on Vapi.ai + ElevenLabs (Custom Voice Cloning, Human-Like Speech, Spam Filter Bypass)
Budget / Salary€300–400
TypeFreelance project
LocationRemote
Posted1 hour ago
Project Description:
We are looking for a specialist to develop and fully configure an AI voice agent for promotional outbound calls in Russia using the Vapi.ai platform integrated with ElevenLabs TTS.
The primary objective is to achieve maximum realism indistinguishable from a live human conversation and eliminate triggers that activate carrier anti-fraud systems and spam detection filters.
Scope of Work:
Custom Voice Cloning/Creation: Create and train a high-quality custom Russian voice in ElevenLabs (Professional Voice Cloning). Using default or public library voices is strictly prohibited. The voice must sound completely natural, including proper intonations, breath sounds, pauses, and micro-accents of live speech.
Scenario Assembly on Vapi.ai: Implement the promotional dialogue logic based on our script (the promotional copy and key objection-handling branches will be provided).
Latency Optimization: Select the optimal model stack (LLM + TTS) and network settings to ensure ultra-low voice-to-voice latency (target: 500–800 ms).
Conversation Flow Fine-Tuning: Configure precise Interruption / Endpointing handling (natural interruptions, waiting thresholds), integrate conversational filler words/reactions, and eliminate unnatural robotic silences.
Anti-Fraud & Spam Filter Adaptation: Optimize the opening hook, cadence, and speech onset timing to prevent telecom carrier bots/answering assistants (e.g., Oleg, Salyut) from flagging the call as an automated robocall.
Candidate Requirements:
Proven hands-on experience with Vapi.ai and ElevenLabs APIs.
Solid understanding of WebRTC, SIP telephony, and conversational flow configuration for Russian-language traffic.
Portfolio of completed cases (audio recordings or interactive demo calls).
Acceptance Criteria:
A live test call demonstrating natural, human-like dialogue flow without robotic artifacts.
The voice agent reliably handles interruptions and performs well under ambient background noise.
The project is fully deployed and handed over to our workspace along with operational setup documentation.
We are looking for a specialist to develop and fully configure an AI voice agent for promotional outbound calls in Russia using the Vapi.ai platform integrated with ElevenLabs TTS.
The primary objective is to achieve maximum realism indistinguishable from a live human conversation and eliminate triggers that activate carrier anti-fraud systems and spam detection filters.
Scope of Work:
Custom Voice Cloning/Creation: Create and train a high-quality custom Russian voice in ElevenLabs (Professional Voice Cloning). Using default or public library voices is strictly prohibited. The voice must sound completely natural, including proper intonations, breath sounds, pauses, and micro-accents of live speech.
Scenario Assembly on Vapi.ai: Implement the promotional dialogue logic based on our script (the promotional copy and key objection-handling branches will be provided).
Latency Optimization: Select the optimal model stack (LLM + TTS) and network settings to ensure ultra-low voice-to-voice latency (target: 500–800 ms).
Conversation Flow Fine-Tuning: Configure precise Interruption / Endpointing handling (natural interruptions, waiting thresholds), integrate conversational filler words/reactions, and eliminate unnatural robotic silences.
Anti-Fraud & Spam Filter Adaptation: Optimize the opening hook, cadence, and speech onset timing to prevent telecom carrier bots/answering assistants (e.g., Oleg, Salyut) from flagging the call as an automated robocall.
Candidate Requirements:
Proven hands-on experience with Vapi.ai and ElevenLabs APIs.
Solid understanding of WebRTC, SIP telephony, and conversational flow configuration for Russian-language traffic.
Portfolio of completed cases (audio recordings or interactive demo calls).
Acceptance Criteria:
A live test call demonstrating natural, human-like dialogue flow without robotic artifacts.
The voice agent reliably handles interruptions and performs well under ambient background noise.
The project is fully deployed and handed over to our workspace along with operational setup documentation.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.