Post reply

Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.

Note: this post will not display until it's been approved by a moderator.

Message icon:

Sum of two plus two?:

shortcuts: hit alt+s to submit/post or alt+p to preview

Topic Summary

Posted by: Springdream
« on: February 02, 2020, 09:13:05 »

I am thinking of a simpler approach: the song is identified correctly even if it starts e.g. 20s after the beginning (because of ads or scpeech at the beginning).
=> a song that has delay 0s (starts right away) is much better than a song that starts later, because you need to listen to the rubish at the beginning
Posted by: Admin
« on: September 11, 2019, 20:19:01 »

There are algorithms to distinguish talk and music, they not very complicated. But Similarity didn't have them and I don't think we add such rarely using feature to it and how users use it, split file to parts?
Posted by: Springdream
« on: January 04, 2019, 10:22:33 »

No reply so far?
Is somebody using radio recordings e.g. from Streamripper?
Radio music files are a good source for free and legal music.
To improve its quality especially ads and wrong starting time needs to be detected.
=> Would it be possible to output the offset (in s) of the pairs? And also allow to automark based on it.
That would allow to detect that music starts later (not at 0s)

Thank you,
Posted by: Springdream
« on: September 16, 2018, 09:15:51 »

currently I get all my songs from radio. I use streamwriter to continuously write e.g. 40 channels (italian, frensh,top100...) to disk. Within a week you get 100.000 new songs or more.

Now, the main challange when it comes to delete duplicates is to delete files that have somebody talking at the beginning or there is another song not yet finished.

=> a new criteria that indicates the correct beginning of a song is helpful, but I am not sure if it is possible to get the information from e.g. Musicbrainz if a song starts later (because of song before) than at 0seconds. Often the song is identified properly even if it begins e.g. 20sec later.

=> an option without internet information could be (it's not done elsewere! so it could be something new): list the delay of all found matches. The file with the smallest delay could be the original (without speach at the beginning) or the file with the least disturbance.

At the end that would allow to retrieve high quality songs even from radio station recordings..