What it does
Editing a video is the fun part. Writing the title, the description and twenty tags afterwards is the part people stop doing, and that is the part the algorithm reads.
This pulls the audio, transcribes it, and uses a language model to write the title, description, tags and thumbnail direction from the actual content of the video, not from the file name. Then it uploads on a schedule and never publishes the same video twice.
Stack
Python, speech to text, LLM, YouTube Data API, GitHub Actions.