Synchronize subtitle file based on audio fragments
€5000-10000 EUR
Geschlossen
Veröffentlicht vor fast 9 Jahren
€5000-10000 EUR
Bezahlt bei Lieferung
I am working on a project that includes the automated download, synchronization and cleaning of subtitles for any given movie (video) file. I already built the automated download of the subtitles and cleaning of the (removing credits etc), but I am a bit stuck on the synchronization part.
In my current approach I have the following:
1) I download subtitles based on filename of the video file
2) I extract audio fragments from the video file
3) *This is what is missing* I want to calculate the offset between the spoken words in the audio fragments and the subtitle file
4) I clean the subtitles
For this project I only want to have the functionality of number 3 provided. So the functionality should take 5 audio fragments of 20seconds and the start position of these audio files (so that you exactly know when the the audio fragments in the movie start) and should calculate the difference between the subtitles and the time these sentences are spoken in the movie.
My audio fragments I test with can be found here:
[login to view URL]!dpxWDAzL!4vRoa53qxh-6_oFShmXj0l_QqA6why7ndcAZZb7g1ZE
[login to view URL]!thBiQB4Q!duh7myW-_yyS4Ht3Q0DXfgsSICvXpidac12o_p-ejI0
[login to view URL]!80QQHTQL!jxJRRA82FUOmUsR0HmuG2umKRAsIQ-o0m1iB_bdLWYA
I know this is something quite hard to accomplish, hence the high value of this project. If you can convince me with the successrate of the functionality, it's more than worth it to me.
Hi, I worked on detecting silence between words in audio to synchronize text with audio, word starts after silence etc... so I can try to solve your problem.
you want subtitle with accurate text timing as final file right??
Hello,
What an interesting project!
My initial approach would be to apply an audio segmentation algorithm to isolate the dialogue and create a timeline pattern of speech. This pattern would simply indicate on the timeline where speech occurs, but wouldn't know the content of the speech itself. The corresponding pattern for the subtitles is then easily generated from the file as it already contains the necessary metadata. These two patterns could then be compared against each other to produce the best matching offset with consideration for video frame rate. This comparison would only look at the start of dialogue blocks, as the rate of speech would complicate a more detailed comparison of the patterns.
Please feel free to contact me here if you'd like to discuss your project in more detail.
Best Regards,
Mitch
Hello
I have been working at Samsung R&D Institute Bangladesh as software engineer. I really appreciate research project. As i work with professionals i think can help you. Together we can achieve success. I can bet within a month we will be able to sync audio with subtitle. I have a great idea to calculate the offset between spoken words. If that works it will be too easy.
Lets start. I will be looking forward for your reply.
Regards
Sajib Dey
Hi, don't really need C to make it. with my programs ans my expertise in the audio and video I can synchronize. I have 10 years in audio producing and several years in video edition, I can.