I ran synchronization on some audio files with some good results, but also some bad results. The Length in Windows Explorer keyed me into the files having extra long silence at the end of the files. After removing the silence, the results were good AFAIK. Any ideas why this silence would affect the timings?
I have the same experience, if there are some extra seconds (with silence) at the end of the audio file, Aeneas can’t handle it very well. If you select the option (in the Aeneas wizard) “End of Audio - Yes, detect the lenght automatically” it doesn’t help much. I usually check how many seconds of silence there is approximately and select the option “Yes ignore a fixed number of seconds”, entering the time there. If all the files (or most of them) have about the same number of seconds, then this works well. The problem is with project where each file is different.
Aeneas uses a text-to-speech process to create a waveform of your text. It then compares it to the actual mp3 waveform and tries to match them to find the phrase breaks. If the mp3 has silence at the end, that probably still shows up as some noise in the waveform and Aeneas tries to match it to some text. I would use Audacity to clip of the extra silence to avoid this problem.