TalkGPT is a speech-to-speech AI assistant based on a set of open API services and offers a range of customized features.
Basic structure of TalkGPT
- To use the speech-to-text and chat completion APIs, you'll need an OpenAI account to create an API key in your user page.
- To use the text-to-speech API, you'll need an AWS account to create an IAM account and a set of credentials on the credential management page.
git clone https://github.com/Smilence7/TalkGPT.git
cd ./TalkGPT
- Python 3 is required, and the testing was done on version 3.8.
- Install the required third-party libraries.
pip3 install -r requirements.txt
cp ./config/config.yml.example ./config/config.yml
vim ./config/config.yml
Set your Access Keys in the config file and adjust the properties to suit your needs.
python3 ./src/main.py
- Press and hold
T
on your keyboard to talk. - You will hear the response from your output device in around 1-5 seconds, depending on your network condition.
Just kill it.
You can have a conversation with the application, just like you would on the ChatGPT website, but in the form of voice input and output.
No input/output restriction.
Speak several sentences and the program will respond with a better or more native version. Primarily intended for language improvement purposes.
- speech-to-speech workflow
- Implement GUI for hotkeys binding & configuration
- Packaging
- Locally fine-tuned TTS model
- Show chat history in GUI
- Function: Replay the last response.