ltx 2.3 vs wan 2.2 bernini model?
what's the difference between ltx 2.3 vs wan 2.2 bernini model? are they both v3?
The problem with LTX is that it works but produces ghosting effects, but on the other hand, it allows you to generate videos with higher resolutions and longer durations. Now, Wan Bernini is a full FineTune built on top of Wan 2.2, focused on editing, so in that case, it manages to be more consistent and without ghosting effects. The problem is that its resolution is lower, the FPS is 16, and the duration is also shorter.
also noticed that bernini doesn't keep the identity as well. Is that expected?
also noticed that bernini doesn't keep the identity as well. Is that expected?
One important thing is to maintain the same resolution as the one used in the Bernini training, which in this case is a 640x640 resolution. It won't be square, but the maximum resolution of one of the sides should be that. Bernini is very sensitive to resolution. But I just saw that this was your case. Another thing would be to try describing the details of the face in the prompt. And as you can see, Bernini tends to copy the shape of the head and other characteristics more, so it won't always turn out well.
bernini is weaker