This updated version has a lower training DIM, which essentially allows greater flexibility at the cost of lower detail. Simply, this is a less "baked" version, with otherwise very similar results. Suggested to use at weight 1 (whereas v1 suggested to use at 0.8 or lower).
This updated version has a lower training DIM, which essentially allows greater flexibility at the cost of lower detail. Simply, this is a less "baked" version, with otherwise very similar results. Suggested to use at weight 1 (whereas v1 suggested to use at 0.8 or lower).