A project about learning how to synchronize subtitles in movies using machine learning.
This work approaches the topic of the Automatic Synchronization of a Movie’s Subtitle. Unsynchronized audio and subtitle files are common within streaming media. As subtitles often are an essential part of the viewing experience, this can have large consequences, possibly making the content inaccessible. Detection of the asynchronous subtitle and its synchronization are necessary in this case.
In this work five existing implementations are explored and tested in detail as well as a sixth one based on an existing implementation and modified by us. These implementations rely on artificial intelligence such as machine learning techniques and speech recognition.
You can find more details about our work in the report file and in each folder seperately. Each folder contains more detailed instructions/descriptions.
- A folder to de-synchronize subtitles.
- A folder to evaluate a subtitle.
- A folder with implementations which synchronize subtitles.
- A folder with our results testing the implementation.
- A folder with our work to try to alter one implementation and improve it.
Some demo videos of our LSTM implementation.
| Before Synchronization (minus 3 seconds shift) | After Synchronization |
|---|---|
 |
 |
| Before Synchronization (random shift in each dialogue) | After Synchronization |
|---|---|
 |
 |
