Скачать или смотреть Voice to Voice AI with Python! Wake Word, Whisper, Ollama, TTS – Full Pipeline Demo

In this video, I show you how to build a full speech-to-speech pipeline using Python – great for your next tinkering or maker project!

We start with Porcupine from Picovoice to detect a custom wake word. Once triggered, OpenAI's Whisper transcribes your speech to text, which is then passed to Ollama (running a local LLM like LLaMA or Mistral) as a prompt. The response from the AI is spoken back using a text-to-speech (TTS) engine in Python.

If you're a hobbyist, maker, or just curious about voice interfaces with AI, this one's for you.

🛠️ Tools Used:
Picovoice Porcupine (wake word)
OpenAI Whisper (speech recognition)
Ollama (local LLM)
Python TTS (pyttsx3 / edge-tts / your choice)

Link to code:
https://gist.github.com/jameselsey/87...

👨‍🔧 I’m a hobbyist maker sharing cool projects – subscribe for more voice/AI/electronics experiments!