<p>Lokr uses a fixed trigger word &lt;artists&gt;wlop&lt;/artists&gt; during training, but the trigger word does not need to be explicitly added during actual inference to achieve the corresponding style.</p><p>The first three images displayed on the page are from the training set, but they are generated using non-training captions during inference; the last five images are samples outside the training set, used to demonstrate Lokr's generalization ability.</p>

Files with SHA256: 56261d0846722916507230c94bbb3fb3b6042d2240d76721aa600dcf3b2a088a