Hey everybody! I'm really excited to share my new model with you all!I decided to separate this checkpoint from the rest of my models because it's built in a very different way from the others and I think it's the best model I've made so far. I merged 3 models (<a target="_blank" rel="ugc" href="https://civarchive.com/models/1162518?modelVersionId=1714002">Plant Milk: Walnut</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/2170933/vete-dreamland-ill">Vete ┃ Dreamland ILL</a>, and <a target="_blank" rel="ugc" href="https://civarchive.com/models/2097759?modelVersionId=2373416">Copper Thyme</a>) then fine-tuned that model with 500 images using OneTrainer. It wasn't very stable though so I merged it back in with the model I had trained it on and refined it by merging again with an older checkpoint I had abandoned.The style is very different from what I intended but I really like the results. Comtesse Raffinée translate to English as Refined Countess. I named it that because the style reminded me of Duchesse but it's more elegant, refined and flexible. Although it has a dramatic and cinematic style of its own it still works well with artist and style tags as well as LoRAs.Image Settings<ul><li>Sampler: DPM++ 2M Karras / Euler A/ Restart</li><li>Steps: 25–50 (base), 30–50+ for Hi-Res Fix</li><li>CFG: 3–7</li><li>Clip Skip 2</li></ul>

Comtesse Raffinée

comtesse_ill.safetensors

Raffinée v1.0

Hey everybody! I'm really excited to share my new model with you all!I decided to separate this checkpoint from the rest of my models because it's built in a very different way from the others and I think it's the best model I've made so far. I merged 3 models (<a target="_blank" rel="ugc" href="https://civitai.com/models/1162518?modelVersionId=1714002">Plant Milk: Walnut</a>, <a target="_blank" rel="ugc" href="https://civitai.com/models/2170933/vete-dreamland-ill">Vete ┃ Dreamland ILL</a>, and <a target="_blank" rel="ugc" href="https://civitai.com/models/2097759?modelVersionId=2373416">Copper Thyme</a>) then fine-tuned that model with 500 images using OneTrainer. It wasn't very stable though so I merged it back in with the model I had trained it on and refined it by merging again with an older checkpoint I had abandoned.The style is very different from what I intended but I really like the results. Comtesse Raffinée translate to English as Refined Countess. I named it that because the style reminded me of Duchesse but it's more elegant, refined and flexible. Although it has a dramatic and cinematic style of its own it still works well with artist and style tags as well as LoRAs.I do plan on trying more testing with finetuning and if I can get a good model from that I'll publish it and detail the process in an article.Image Settings<ul><li>Sampler: DPM++ 2M Karras / Euler A/ Restart</li><li>Steps: 25–50 (base), 30–50+ for Hi-Res Fix</li><li>CFG: 3–7</li><li>Clip Skip 2</li></ul>

Files with SHA256: c4f6dcdd1fb484b31a60ad35fd9a2c0b4836777fe2eb938cff6d2a8ddf954517