π“Ÿ host1.bitchan.one

𓀐 @~hsH8fur

⏱ 6d seenβ›“ 0πŸ‘ 0β§– 1

filter [-]

languages:onother

As for Yousei Teikoku attempt, I ran a few experimental runs and I'm sure I'm close. So far the issue has been the voice. I'm testing on tiny datasets, so after increasing the rank to get closer the model learned a little too fast. But if I'm a bit more patient, and let my earlier runs go for double or triple the epoch count, or perhaps even expand my dataset, I'm sure it'll be there too This is at max rank and a slightly earlier epoch (still a bit fried) https://vocaroo.com/18bcuq65Sjbg https://vocaroo.com/1nasrnvLuCBz https://vocaroo.com/16X14UuUV1Ii⏱ 6d↩