I found a bug (i think) in the new 241 version.
When selecting the most accurate (biggest) model, it outputs random garbage text for quite a while, then after a few minutes into the video it starts picking up the correct language, and translates it correctly.
On the default middle setting it translates the video fast and correct!
PS: on the most accurate setting it also often crashes the app at the very end of loading the video, before it begins transcribing it (.mp4 720p vids).
The video file itself is not the issue, since the middle setting does the whole video correctly without crashing.
The program has no error-log when it crashes, so I cant be helpful in providing that info, so maybe an idea to implement such a log in the future? (see image below for results)