🧸 Krea 2 Clay Figurines turns one subject into a cute handmade clay figurine: soft matte modeling clay, rounded shapes, thumbprints and small sculpting marks, pastel colors, shot like a toy photographed on a plain backdrop.
✨ Trigger word: clymworld (put it at the start of the prompt)
🧩 Base model: Krea 2 (tested on Krea 2 Turbo; trained on Krea 2 Raw)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🎯 What it is good at: ONE subject on a plain pastel backdrop
Animals, birds, flowers, everyday objects, sweets, a single building or vehicle. Prompt pattern that worked:
clymworld, <one subject>, <view>, on a plain <color> pastel backdrop
🖼️ Example prompts (all from the gallery, unedited)
• clymworld, a small brown bear cub sitting, front view, on a plain mint pastel backdrop
• clymworld, a white duckling with orange feet standing, side view, on a plain pink pastel backdrop
• clymworld, three round donuts with pink, white and chocolate icing on a small plate, three-quarter view, on a plain blue pastel backdrop
• clymworld, a tall pastel purple wizard tower with a pointed roof, front view, on a plain pink pastel backdrop
• clymworld, a small yellow school bus with round wheels, side view, on a plain cream pastel backdrop
Short plain prompts like these already work; you do not need a long style paragraph.
⚙️ Settings that worked for me (Turbo)
• LoRA strength 1.0
• 10 steps, CFG 1, sampler euler, scheduler simple
• Size tested: 1088 x 1920 (portrait)
• I did not test other strengths, other sizes or the Raw model with the final file.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 What my tests showed (honest numbers)
• Test 1, 10 mixed prompts (7 single subjects + 3 wide scenes: a city street, a kitchen, a mountain lake): 5/10 looked like real clay. Single subjects 5/7, wide scenes 0/3.
• Test 2, 10 NEW single-subject prompts (not seen before, written and locked before generating, includes two hard ones: shiny bananas and a bus with windows): 10/10 looked like clay.
• The same prompts without the LoRA: 0/10 looked like clay (photos or 3D renders).
• "Looked like clay" = the whole picture is matte clay with sculpting marks, no mixed shiny / wood / plastic look, no face on anything that is not an animal. I judged every image at full size myself. 10 images is a small sample: expect most, not all, of your results to be clean.
⚠️ Known limits
• Wide scenes do not work. Streets, rooms and landscapes come out as a "clay-look CG render" (glowing lights, glossy water, faces on rocks and cars), not as clay. Use this LoRA for one subject on a plain backdrop.
• Some prompts give a shiny or wooden surface. In test 1 a red apple came out glossy like a photo and a toy train got wood-grain parts. Try another seed or write "matte clay" for the object.
• Animals get small dot eyes; objects stay plain without faces.
• Small training set, one look: it is a style, not a guarantee for every subject.
• No people were used in training or in my tests.
🛠️ About the training
• 20 synthetic images (animals, birds, flowers, objects, food, small buildings) generated with Krea 2 by me, each with a written caption; 4 more were held out for testing. No scraped, third-party or real-person images.
• LoRA rank 32 on Krea 2 Raw, 960 training steps, trained locally with musubi-tuner. File: checkpoint at step 960.
🔗 More from me (same base model, Krea 2)
• Krea 2 Origami World: the same idea in folded paper (https://civitai.com/models/3019950)
• See also my other Krea 2 LoRAs in the Suggested Resources on this page.
📝 Notes
All gallery images are AI-generated with this LoRA (Krea 2 Turbo, strength 1.0), not edited, with their prompts and seeds in the image metadata.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⭐ Found this useful?
• Like if it saved you time
• Comment your results, I read every one
• Follow for new ComfyUI workflows, all tested on 16 GB VRAM
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📜 Licence and notice
Krea 2 is licensed under the Krea 2 Community License Agreement. For more information, visit https://krea.ai/krea-2-licensing.
This LoRA is a modified version (a Derivative) of the Krea 2 model. It is not an official Krea product and is not endorsed by Krea. Use of this LoRA and of the images it makes is subject to that Agreement and to the Krea Acceptable Use Policy (https://krea.ai/krea-2-use-policy). Commercial use is only allowed under the Agreement's company-wide revenue limit of US$1,000,000 per year (section 2.3).