imputnet / cobalt

save what you love
https://cobalt.tools
GNU Affero General Public License v3.0
10.46k stars 739 forks source link

Download partial audio using a start and finish delimiters #442

Open maherbel opened 2 months ago

maherbel commented 2 months ago

describe the feature you'd like to see Download a specific portion of a given youtube video's audio. This specific part is delimited by start and end.

describe alternatives you've considered I considered downloading a "portion" of a video to extract the audio but that is not possible right now.

additional context I've build a summarisation pipeline that takes as an input a YouTube url (short or video) and summarises it. To do so, I actually use Cobalt's API to request the video audio's and pass the computed audio url to a "speech-to-text" component (built on AssemblyAI api). The issue here is that for any video that is longer than a few minutes, the audio file starts to be pretty relatively big for the downstream speech to text component. This brings me to the current feature request that would help me divide (and conquer) through multithreading the computation of each small part of the audio into small parts of text that I would merge.

lostdusty commented 2 months ago

I believe its being added here: #350

maherbel commented 2 months ago

Great to know ! Seems like it should implement exactly what Im looking for 🤩