Pod2Reels

This is a side project that converts youtube podcasts into smaller chunks of reels. This is how it works:

1. Download the youtube podcast
2. Get the transcript of the podcast
3. Use GPT-4 api to convert the transcript into several categories based on the content discussed in the podcast. This also provides the starting and ending timestamps of the content for each category.
4. Use ffmpeg to split the podcast video into several chunks based on the timestamps.
5. Generate subtitles for each chunk using ffmpeg.

Outputs

For the input video from the Lex Fridman podcast, about 10 reels are generated with subtitles embedded in the mobile sized reels. The following shows the input video and the output reels(only 2 shown here).