Training new vocals now that I have a 400k-steps checkpoint.

[blizzard090]

hello all,
I trained 3 singers in one model and trained at 400k steps (G_400k and D_400k), the results were quite good.
Now I plan to train 3 more singers.
With this new plan, I should use the pre-trained model: G_400k and D_400k or G_0 and D_0. Which one will be better?
It is assumed that the new singers are similar in gender and music genre to the old singers.