You have to find a noise near the beginning. Visual clues such as a door slamming are far far easier than someone speaking. Really hard sometimes to look at someone talking and work out if it needs to go forwards or backwards! There wasn't much in this so I had to go with the girl talking at 2m07s, but the audio comes in about 0.1 seconds after the scene cuts to her.Night457 wrote: ↑Wed Jan 04, 2023 3:24 pm Thank you David32441. This looks to be a rough watch, I will have to give myself some time to prepare for it. I feel your pain for the audio manipulation and retiming. I was never successful attempting that sort of thing with Audacity. Hurrah!
Thank you kal for the eMule share, and congratulations on your first post!
The one near the beginning you can fix the timing by experimenting with "mkvmerges" audio delay, in this case 1700ms (1.7s). That gives you the start. Then I used Audacity's "Change tempo" it has a "percent change" number and a handy "Length (seconds)" which I experimented with increasing from 2014.07 -> 2017.4 (33m 37s). Then very importantly, crop that same number of seconds off the audio file at the end (usually silent), or I think mkvmerge will get confused by having an audio track run longer than the video and might try and change the audio speed to match the video length which will undo your changes. Then merge the wav audio as a new track into mkvmerge and test it. Check your audio near the beginning and a audio bit near the end. For me that was a clicking lighter at about 17m and a hammer and nail near the end
Then if it's out by say 0.5 seconds you go back to your "original" wav file and repeat, but add a tiny fraction more % until the length says 2017.9 and keep repeating until it's as near to perfect as you can be bothered!
When you're finally happy with the -1700ms - then take that off the wav file. Then encode that. I use RazorLame at 128kb. Make sure your new audio track is the default audio track, the old one is default=no, and you've added the subtitles!