Project profile
Scope, role and technology
Visual evidence
PBX configuration, n8n workflow views, transcription output, audio demonstrations, API/intent routing, and TTS experiments from the BAYNAO voice R&D system.
BAYNAO voice automation experiment
This project belongs to the BAYNAO ecosystem. It connects n8n to the local Issabel PBX/server so a real telephony flow can record caller audio, send it into automation, transcribe it, route the request into business/AI logic, and experiment with generated voice responses.
Issabel call flow
- Issabel extension used for controlled testing.
- Audio recorded as a WAV artifact.
- Webhook/curl handoff sends the recording to n8n.
Speech-to-text
Local Spanish transcription with whisper.cpp was tested successfully so recorded audio can become text for downstream BAYNAO workflows without requiring one external provider for the full voice stack.
Text-to-speech experiments
Response generation has been tested with ElevenLabs and explored with a local Piper approach. This remains R&D; real-time interruption handling and latency are not presented as solved production capabilities.
Architecture
What this prototype demonstrates
- Hands-on integration across PBX, audio files, webhooks, STT, workflow orchestration, business APIs and TTS.
- Keep STT/TTS providers replaceable.
- Do not label a turn-based prototype as a fully real-time conversational agent until latency and interruption behavior are solved.