<h3 id="moss">MOSS</h3>Still experimenting with moss training... Showcase images :<ul><li>SFT model + turbo lora 4 steps</li><li>Euler simple / 5 steps / CFG 1</li><li>1024x1536</li><li>Lora strength : 1</li></ul><h3 id="krea-2">KREA 2</h3>Showcase images :<ul><li>Turbo model</li><li>Euler / simple / 8 steps / CFG 1</li><li>768x1152</li><li>Lora strength of 1.3</li></ul><h3 id="ideogram-4">IDEOGRAM 4</h3>Used json captioning with art_style : "Studio Ghibli style" and medium : "anime_screencap"v2: Try other settings and the style seems a bit better with this onev1: First training on Ideogram4<h3 id="anima">ANIMA</h3>First Lora trained on Anima, and it seems really good for a first try.Showcase images :<ul><li>er_sde / simple / 30 steps / CFG 5</li><li>768x1152</li><li>Lora strength of 1</li></ul><h3 id="klein-9b">KLEIN 9B</h3>v1 :Showcase images were generated with the base model, but I've posted same prompt images with the distilled one which are way better imo. None of the images were upscaled.<h3 id="zib">ZIB</h3>v1 :First time training a Z-image Base lora, so still experimenting. Showcase images :<ul><li>Euler / normal / 20 steps / CFG 5</li><li>Hires steps : 20 / NMKD_SIAX @0.2-0.5 / x1.5-x2</li><li>768x1152</li><li>Lora strength of 1</li></ul><h3 id="qwen">QWEN</h3>v3 :After training way more qwen loras, I've tried to apply everything I've learned with this one. Trigger word is now simply 'ghibli style'It should be more flexible and closer to the style of the movie. But to do characters of the movie, v2 is better. v2 : I've experimented way more with qwen training and this version is way better than v1. The lora is a bit big (600mo) because it's like 18k steps. And I've added a trigger token for the style : 'KikiLaPetiteSorciere style' To generate all images, I've used Qwen-Image (BF16) + <a target="_blank" rel="ugc" href="https://civarchive.com/models/1854805?modelVersionId=2235524">Lightning Lora (4 steps) </a>+ my lora with weight of 1. DDIM with Beta, 4 steps, CFG 1.For some of them, I've upscaled with my flux lora version of this one, and <a target="_blank" rel="ugc" href="https://tensor.art/workflows/920224064797170426">this workflow.</a>v1 :First qwen-image lora i've trained. I'm still testing and learning<h3 id="flux">FLUX</h3>First attempt at a Flux Lora (trained on Civitai).The dataset consists of 28 screencaps (1920 × 1024) from the Ghibli movie Kiki's Delivery Service by Hayao Miyazaki.I used Florence2 to caption the images, then manually corrected the captions to fix any accuracy errors and to add the trigger phrase:'a KikiLaPetiteSorciere style image of'<h3 id="sdxl">SDXL</h3>Trigger words : KikiLaPetiteSorciere styleFor the character Kiki : kiki (majo no takkyuubin), red bowTrained on 1000 images screencap from the movie Kiki's Delivery Service by Hayao MiyazakiAll images displayed were made with Dreamshaper XL Lightning :Lora's strength of 11216x832CFG 26 stepsDPM++ SDE KarrasClip skip 2Only the jedi one was inpainted (she had 3 legs...)

Ghibli style (Kiki's Delivery Service)

ghibli_style_moss.safetensors

Moss v1

Krea2 v1

Ideogram4 v2

Ideogram4 v1

Anima v1

Klein 9B v1

ZIB v1

Qwen v3

Qwen v2

Qwen v1

Flux v1

SDXL v1

Files with SHA256: 0f492ae0691d2d243cb4a261f9935f4a18023a33f7ec98425d3c75e8c620f919