Voice-Cloned TTS Phone Assistant

I need a text-to-speech system that can learn an exact voice from a pre-recorded sample and then speak in that voice during live telephone conversations. The primary purpose is a personal-assistant style companion, so the speech must sound natural and respond with minimal lag while a call is in progress.

After I supply the reference recording, the program should train itself to match tone, pitch, and cadence, then expose a simple way—GUI, CLI, or REST endpoint—to feed it text in real time. That audio stream has to be injected straight into an active call through a standard telephony interface such as SIP, Twilio, or a comparable VoIP gateway; if you know a better route, let me know.

Platform isn’t locked down, but Windows support is essential and cross-platform flexibility is a bonus. Low latency, clear audio, and stability during long calls are critical acceptance points.

Deliverables:
• Voice-cloning TTS engine trained on the provided sample
• Live-call integration module ready for SIP/Twilio routing
• Test harness (simple GUI or API) to type text and hear it in-call
• Deployment guide and fully commented source code

Once everything is running smoothly, I will handle the conversational logic myself; your focus is the voice and the phone line.