AI assisted transcription and subtitle creation

[lsg1310]

I do have a RTX 4060 so I'll see if I can try out that engine. Is large-v3 any better than large-v2?

[ghost]

Is large-v3 any better than large-v2?

v3 is the newest model. I personally stil use v2, because it still gives me better results. Just try both so you can compare.

[Night457]

v3 is the newest model. I personally stil use v2, because it still gives me better results.

...and this is one part I got wrong.

[pillowbaker]

Is large-v3 any better than large-v2?

v3 is the newest model. I personally stil use v2, because it still gives me better results. Just try both so you can compare.

When I am going in-depth with my subtitles, I do whisper transcriptions using both v2 and v3 models. I have seen a number of people online saying the v2 is still more reliable, but occasionally v3 will catch something that the other misses. I am interested in hearing about other whisper/translation techniques and sites that users have been experimenting with.

[lsg1310]

I wish there was just one clearly superior method since I'm so lazy

[lsg1310]

Does it just use Google Translate when you check 'Translate to English'?

[DreamScape]

Does it just use Google Translate when you check 'Translate to English'?

No. It will actually use the Whisper model to output in English without ever generating the subtitle in the original language.

The result will be different from translating as a second step after the subtitles are created.

[lsg1310]

Interesting. Guess I'll have to compare and contrast