<p>Check out new LoRA <a rel="ugc" href="https://civarchive.com/models/1564264/orange-mood">Orange Mood</a>, loosely based on Denslow's Mother Goose</p><p></p><p>This LoRA based on illustration by William Wallace Denslow. Particularly on Denslow's Mother Goose and The Wonderful Wizard of Oz.</p><h2 id="read-before-doing-anything-nsfw-ug9qtr9fk">Read before doing anything NSFW</h2><p>Most illustrations used in dataset depict children, so this LoRA tries to depict children or at least add some childlike features. Even with negatives characters, especially female, will drift towards being childlike.</p><h2 id="strength-70g9zln7e">Strength</h2><p>Sensible to use from 0.7 to 1.5. Further it exaggerates too much.</p><h2 id="composition-and-background-yc63gavlc"><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/358e3438-2437-4f6c-b5d4-0db23e2eeed8/width=525/358e3438-2437-4f6c-b5d4-0db23e2eeed8.jpeg" />Composition and background</h2><p>Most illustrations have lots of empty space and simple backgrounds. It's a feature of Denslow's illustrations.</p><h2 id="feedback-needed-vjuuco1ca">Feedback needed</h2><p>Have fun generating. Don't forget to post some pics and share your experience. I can only do so much testing myself and sure there is lots of room for improvement.</p>

Inspired by W. W. Denslow's Illustrations

<p>Retagged and somewhat cropped all images, reduced amount of The Wizard of Oz, since I like Mother Goose style more. It is now not so rich on children, so it would be somewhat easier to tag against. Still, most characters in dataset are children, so chars will either have childlike features, or resemble few old people in dataset. Choice is yours.</p><p>This one have HUGE problem - it might be overly generous with signatures and other markings, but since they are all on empty plain background, should be pretty easy to remove.</p>

Files with SHA256: 421c14943a9585498ba423542c382b0a68b30e170b254b803ce9e55d52f936fe