Pycasts is a reference API for producing long form spoken audio and scripts for podcasts, powered by Nitric, Suno Bark and Llama 3.2.
This project uses also uses smaller LLMs for producing podcast scripts, which is baked directly into the resulting container when we deploy, via the models directory of the project.
This will start the project on your local machine, and you will be able to use the provided local nitric dashboard, or your HTTP client of choice to interact with the API.
Audio models are downloaded ahead of time using the API to be stored locally in a bucket, this can be done by hitting the /download-model endpoint:
Watch the logs in your CLI to see progress for now. Your output audio will be available in the clips bucket once everything has generated. See the local nitric dashboard storage, to download your finished podcast.