Is it possible to transcribe podcasts by URL?

[fearby]

Is this likely to be possible - using ts can do it if I download a podcast as an mp3 but would be good if it was possible to extract summaries like in YouTube.

[xeniode]

You would need some tool to grab the audio at least. I have integrated yt-dlp into fabric for much more functionality with youtube videos. If you use that, you can grab audio and pipe it through ts.

yt-dlp -x --audio-format mp3 https://www.youtube.com/watch?v=dQw4w9WgXcQ

[djl0]

Just to clarify, the fabric-official yt tool only uses the youtube-provided transcript, right?

fabric/installer/client/cli/yt.py

Line 98 in d75ea47

transcript_list = YouTubeTranscriptApi.get_transcript(video_id, languages=[options.lang])

How do you tell if it's auto-generated or user-provided transcript? And are you saying that the fabric transcription would be better than youtube's autogenerated?

Thanks

[fearby]

You would need some tool to grab the audio at least. I have integrated yt-dlp into fabric for much more functionality with youtube videos. If you use that, you can grab audio and pipe it through ts.

yt-dlp -x --audio-format mp3 https://www.youtube.com/watch?v=dQw4w9WgXcQ

Can I ask how you did this? Thanks

[djl0]

@fearby if the podcast is on youtube, you can use the command Xeniode posted above. If not, you need the mp3 link from either rss feed or the webpage or something. After downloading the audio, you would use the ts command in this repo to transcribe the audio to text. then pipe that text into fabric.

If the podcast is on youtube, i'm generally fine with the youtube transcription (which I imagine is lesser quality), so i just pull the transcript itself yt-dlp --quiet --no-warnings --skip-download --write-subs --write-auto-subs --sub-lang en --sub-format ttml --convert-subs srt --output "%(title)s-%(uploader)s--%(upload_date>%Y-%m-%d)s.%(ext)s" --restrict-filenames <youtube video url or id> then i use sed to clean up the srt file (i can share that if you're going that route, but i don't want to crowd this up too much).

[ujconsulting]

You would need some tool to grab the audio at least. I have integrated yt-dlp into fabric for much more functionality with youtube videos. If you use that, you can grab audio and pipe it through ts.

yt-dlp -x --audio-format mp3 https://www.youtube.com/watch?v=dQw4w9WgXcQ
how did you install/ integrate yt-dlp into fabric? For Example how did i get path plus filename to the ts script as an option, so that it downloads the audiofile, passes it to ts, transcribe it and save the transcript somewhere

[drlakshan]

@fearby if the podcast is on youtube, you can use the command Xeniode posted above. If not, you need the mp3 link from either rss feed or the webpage or something. After downloading the audio, you would use the ts command in this repo to transcribe the audio to text. then pipe that text into fabric.

If the podcast is on youtube, i'm generally fine with the youtube transcription (which I imagine is lesser quality), so i just pull the transcript itself yt-dlp --quiet --no-warnings --skip-download --write-subs --write-auto-subs --sub-lang en --sub-format ttml --convert-subs srt --output "%(title)s-%(uploader)s--%(upload_date>%Y-%m-%d)s.%(ext)s" --restrict-filenames <youtube video url or id> then i use sed to clean up the srt file (i can share that if you're going that route, but i don't want to crowd this up too much).

HI there can you please teach me how to use ts in fabric? Thanks

[nic0711]

How about a link to an Apple Podcast (they normaly have a transcription) or a direct MP3 file link from podigee.io?