Bland AI built something twice as good as this - I've gone pro and joined them in their mission of making the best AI Phone integrations possible. Now it's approaching 10, 100 times better? Maybe a thousand?
Anyways, [https://bland.ai] is insanely useful and we're working every single day to build the best calling systems that apps can run on. They've even got a startup program.
Go there!
The Original, personal pet project README inside
Call (844) 598-0141 to try it out!
AIPhone is an AI-based phone assistant that provides troubleshooting suggestions for various issues. The application is built using Flask, the Twilio API, and the OpenAI GPT-4 model. AIPhone is capable of understanding user speech input and providing appropriate responses to help them resolve their problems.
The application is available at (844) 598-0141. It's still under development, so if you encounter any issues, please report them here or through the contact section on the homepage.
AIPhone is designed to handle incoming phone calls, understand the user's speech input, and provide troubleshooting suggestions based on the user's questions. The application maintains a conversation history to provide context-aware responses, making it more effective in assisting users.
- Utilizing the GPT-4 model to provide intelligent and context-aware troubleshooting suggestions.
- Maintaining conversation history for a better understanding of user issues and providing relevant responses.
- Handling speech recognition and text-to-speech using the Twilio API.
- Conversation history
In future iterations of AIPhone, I plan to add the following features:
- Much more expressive voices using Elevenlabs text-to-speech API
- Fix delay after user talks
- Smarter AI model - currently on GPT-3.5-turbo for performance, using GPT-4 caused timeouts because it takes longer to process
- Web Portal
- Integration with aiitsupport.net for business-specific troubleshooting and documentation
- Hand-off or automatic hangup when call is over
While building AIPhone, I worked with the following technologies and tools:
- Flask (Python web framework)
- Twilio API (Phone call handling, speech recognition, and text-to-speech)
- OpenAI GPT-3.5 and GPT-4 (AI language model)
- Heroku (Deployment platform)
I also used other tech before it was working on the phone:
- Elevenlabs API and custom voices (I cloned my own voice!)
- tkinter for GUI
- PyDub, PyAudio for dealing with audio files
- OpenAI Whisper API for transcription
Building AIPhone has been a valuable learning experience. I've never built something before that works over the phone! This project involved working with various technologies and tools, such as Flask, Twilio API, OpenAI GPT-4, and Heroku. The development process involved gaining a deeper understanding of speech recognition, text-to-speech, and AI language models. Additionally, it provided an opportunity to work with a deployment platform to make the application accessible to users.
The AIPhone project has been an exciting journey, and I'm glad that I can work with GPT technology in it's infancy while also having the skill to integrate it into something as unique as a phone call. It's a much different environment than web development but I like challenges. Thanks for reading!