Earkind

Earkind is an AI tool that generates podcasts using neural expressive text-to-speech and programmatic audio editing. With Earkind, you can create a full podcast episode and description based on a list of news articles and research papers. Simply provide a .txt file with a selection of plain text and a list of recent Arxiv papers as URLs. Earkind crawls the Arxiv database, extracts information from the raw PDF text, and utilizes the ChatGPT API to generate content.

One of the unique features of Earkind is the ability to define characters for the podcast. Hosted by Giovanni Pete Tizzano, an overly hyped and earnest tech bro, along with sarcastic analyst Robert and witty research expert Belinda, the podcast becomes an organized conversation between these characters. Scripts are created for each section, and the program parses them accordingly.

Once the recordings are done, it’s time for editing. Earkind offers a variety of jingles, sound effects, and background music to enhance the podcast. Using Pydub, the tool automatically adjusts volumes, overlays sections, and combines narrations and audios.

In addition to the podcast itself, Earkind also generates timestamps and descriptions for each episode. This makes it convenient for listeners to navigate through the content. The titles and overall description are generated by ChatGPT as well.

Earkind is an innovative tool that has the potential to create personalized audio content. The team behind Earkind is eager to receive feedback and suggestions from users, and there are plans to make the code public while providing an in-depth explanation of the design thinking behind the tool.