Trained on images from the movie <a target="_blank" rel="ugc" href="https://www.imdb.com/title/tt2576852/?ref_=nv_sr_srsg_0_tt_8_nm_0_in_0_q_kaguya">The tale of the princess Kaguya</a> by Isao Takahata<h3 id="qwen">Qwen</h3>2512It was trained on 2512, but I prefer to use the original Qwen-Image model for inference.Showcase images :<ul><li>Original Qwen-Image (BF16)</li><li>Lightning lora (4 steps)</li><li>5 steps / 1 CFG</li><li>DDIM &amp; Beta</li><li>Lora strength 1</li></ul><h3 id="flux">Flux</h3>v2Fixed the straigth line issue with a bigger and better dataset. It seems better at following prompt with this version, but the style is a bit different from v1, colors are more vibrant.Maybe adding "pastel colors watercolor" can help having less vibrant colors.v1I may retrain the lora one day because there is often a straight vertical line appearing (going to the sun/sky/moon) :<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c3bb0531-598d-4593-8fee-5951315b3fb9/width=525/c3bb0531-598d-4593-8fee-5951315b3fb9.jpeg" />

The Tale of the Princess Kaguya style (Takahata's Ghibli movie)

Better at prompt following + more diverse + fixed the straight line issue

Files with SHA256: ddd938eea8b40d017f432bb99dd2582fbb212d5f63224b4fd405272589d8ca83