Style Lora of <a target="_blank" rel="ugc" href="https://www.pixiv.net/users/99130349">这也是QAQ - pixiv</a>Note: Every major version has a different style (datasets also do not overlap), so choose the one that best fits your needs.

QAQ / Style

QAQv4.safetensors

QAQ / Style V4.0

v4.0 vpred

v5.2 IL

v5.1 vpred

v2.2 vpred

v3.2 vpred

v5.0 vpred

v3.1 vpred

v3.0 Vpred

v3.0

v2.1

v2.0

v1.2

v1.1

v1.0

Style Lora of pixiv QAQv1.1 update : organize dataset, optimize learning rate and increase total learning steps.v1.2 update : organize dataset(deleted his old style images)v2.0 added : little different style. can use with v1.2 because the dataset does not overlapv2.1 update : orgainize datasetv3.0 added : another different style. dataset does not overlap with other versionsv3.0 vpred added : the dataset is almost identical. I tried training a LoKr (Lycoris) based on a noobai-vpred model<ul><li>11.12 : replaced the file with an improved version</li></ul>

Files with SHA256: f736a5c3184c63aff97741fa216a8174179f6acd4b940b57bbf407b2e7ec8d91