I was so pissed at how slow Larry Summers spoke on the @theallinpod i cancelled all my Saturday plans and wrote an app that adjusts the speed of individual speakers in a podcast.
Try it out, link in replies
Try it out, link in replies
It uses assembly ai to identify speakers and segments, then bun.spawn to use ffmpeg and rubberband to adjust the speech and deconstruct/reconstruct the audiofile
Larry Summers spoke at a grueling 120 wpm, while the others spoke at 206, 186, 199 and 186 wpm, a single scalar value to adjust speed was simply inadequate, at 2x it put Larry barely above Chamath at 1x
Will add: speaker preview, multi file processing frontend (backend can handle it already) with jobid queries, will consider dockerizing and hosting it, if i do ill also add tagging with podcast name and cacheing
• • •
Missing some Tweet in this thread? You can try to
force a refresh
