Hey GPT!
Models used:
- ChatGPT Whisper: for speech recognization, used in
recorder.py
ChatGPT API Used:
- Speech :
https://api.openai.com/v1/audio/speech
for text to voice, uses tts-1 model - Completion :
https://api.openai.com/v1/chat/completions
for completions, uses gpt-3.5-turbo model
I'm currently working on an ongoing project to develop my own voice-activated device, similar to Google Home or Alexa.
At the moment, it can identify a specific wake-up phrase, like "Hey GPT!", and then respond to voice commands. The speech recognition is performed locally using OpenAI Whisper. Once a command is detected, it's sent to the ChatGPT API for processing, and the response is generated using the OpenAI Text-to-Speech API.
To set it up, you'll need to use the init.py script and repeat your chosen wake-up phrase 10 times to initialize the device. After that, you can use the recognize.py script to activate the device with the wake-up phrase and then issue commands.
$ export OPENAI_API_KEY=YOUR_API_KEY
$ pip install -r requirements.txt
$ python3 init.py
$ python3 recognize.py