The Frontier Today is a daily news show focused on the bleeding edge of emerging technology. Consistent with the theme, the show is algorithmically generated using State of the Art (SotA) consumer and open-source AI technology. The show and its associated code is built and supported by myself, Justin Hutchens -- AKA "Hutch". The show is created daily using custom Python code and a number of different technologies to build curated news stories. The created episodes include speech that is tailored to my own writing (in both tone and sentiment), and even my own voice (through AI voice cloning technology). All of this was accomplished using leading open-source and consumer AI. The technology stack included the following:
OpenAI GPT Model(s) - OpenAI API was used to create a custom language model to resemble my own writing style. Used a series of in-context learning techniques to train a Large Language Model (LLM) system on my own style of writing, to generate content consistent with my own tone and sentiment.
ElevenLabs (Custom Voice Model) - ElevenLabs is a leading AI company focused on text-to-speech and voice-cloning models. I leveraged the ElevenLabs service to train a professional voice model on 3-hours of captured audio of my voice.
Requests - Requests is a commonly used library in Python that can be used for interacting with REST APIs. This is one of the tools that was used for interacting with news sources.
Selenium - For news sources that do not have an official REST API, Selenium was used as an alternative. Selenium is a browser driver utility, which allows you to use an actual browser to execute requests (in the same way that an actual desktop user would), by using an actual web browser. In my case, I used the Mozilla FireFox browser with the Geckodriver.
Beautiful Soup 4 (BS4) - BS4 is a library that is using for parsing structured text like HTML. This is used in conjunction with Selenium to extract meaningful news stories from HTML source code.
MoviePy - MoviePy is a library used for algorithmic audio editing and processing. For this project, this library has been used to stitch together the generated audio segments, transitions, and intro/outro clips into a single cohesive and polished podcast episode.
BuzzSprout API - The BuzzSprout API is used to automate the upload, hosting, and RSS listing of episodes.
Blogger - The Blogger API is used to automate the hosting and updating of the The Frontier Today website.
NameCheap - Domain name registrar used for domain registration and DNS management for hosting.
