AniSee

LoKr merge on anima-base-v1.0 — the same model the LoKr was trained on.The most neutral and most flexible variant: the widest range of subjects and styles, and the plainest default look. If you want to drive artist tags or push your own style hard, start here.40 steps, CFG 4–5, sampler <code>er_sde</code> / <code>simple</code>.This is also the right foundation if you want to train your own LoRAs on AniSee — same architecture as Anima-Base, and CircleStone Labs recommends training LoRAs on the Base variant in every case.Not an all-in-one file — text encoder and VAE separately.

Files with SHA256: bc02a1da5db1317d0c3c04f5c4c96768f01cd5a2d6e97ba60f9135a83baf2387