This is a LoRA of the internet celebrity Belle Delphine for Stable Diffusion XL. (For my previous LoRA for 1.5 based checkpoints see <a target="_blank" rel="ugc" href="https://civarchive.com/models/38290/belle-delphine">here</a>.) This LoRA is quite flexible, but this should be mostly thanks to SDXL, not really my specific training.As trigger word "Belle Delphine" is used. Additionally, “braces” has been tagged a few times in the source images, so you might also be able to use this as a tag during generation.I suggest a default LoRA weight of 1.I tested this LoRA with the base model + refiner, but I wasn’t really happy with that, as the refiner removed characteristics again.Most images were generated with <a target="_blank" rel="ugc" href="https://civarchive.com/models/112902/dreamshaper-xl10">DreamShaper XL1.0 Alpha2</a>.Additionally tested were <a target="_blank" rel="ugc" href="https://civarchive.com/models/114238/tdg8uus-sdxl10-mix">Tdg8uU's SDXL1.0 mix</a> and <a rel="ugc" href="https://civarchive.com/models/117259/anime-art-diffusion-xl">Anime Art Diffusion XL</a> (the last 4 wide images are 4 different models). These images have been rendered using <a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI">ComfyUI</a> using quite a high-fidelity workflow (also depending on the image). But you can assume results were produced with a 1024x1024 base resolution (with different aspect ratio), then latent based img2img upscale with a factor of 1.5 and then also using ComfyUI-Impact-Pack for face details (if needed). The workflow should be embedded with the example images. It should work at &gt;= 8GB VRAM. To make use of it you will need the following custom ComfyUI nodes:<ul><li><a target="_blank" rel="ugc" href="https://github.com/ltdrdata/ComfyUI-Impact-Pack">ComfyUI-Impact-Pack</a></li><li><a target="_blank" rel="ugc" href="https://github.com/Ser-Hilary/SDXL_sizing">Automatic Sizing for SDXL</a></li><li><a target="_blank" rel="ugc" href="https://github.com/WASasquatch/was-node-suite-comfyui">WAS Node Suite</a> (only really used to save images with a timestamp, besides that not used)</li></ul> <h3 id="heading-96">Training:</h3>I assume some people might be interested in how this was trained, but I will have to disappoint you, as it was my first try with default settings and I am sufficiently happy with it, so no secret tips. In detail I used the following <a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss/blob/278b6937a09f5133d4b2a6b19145ad7c43135d49/presets/lora/SDXL%20-%20LoRA%20adafactor%20v1.0.json">base configuration file</a>.With those settings I used 151 uncropped, manually captioned high resolution images with bucketing in the resolution range 512-2048, 32 as bucket steps. The image dataset was set for 25 repetitions, additionally there was a regularization dataset (freshly generated with SDXL) with about 3000 images, set to 2 repetitions. Originally, I set training to 30 epochs (as that was the default of the previous mentioned config file), but around 8 epochs the results began to overbake. For that reason I stopped after 18 epochs (around 27000 steps). The final LoRA is then a merge of epoch 3 (10%), 6 (70%), 9 (12%) and 18 (8%). After that I resized it with a factor of 0.96, target rank 32 resulting in the ~40MB file. As you can see in the percentages (of the epochs in the merge), good LoRA results were already achieved relatively early.Training was done on a rented 3090 (cloud). Should you have any additional questions, feel free to ask.

Belle Delphine XL

DI_belle_delphine_sdxl_v1.safetensors

v1.0

This is a LoRA of the internet celebrity Belle Delphine for Stable Diffusion XL. (For my previous LoRA for 1.5 based checkpoints see <a target="_blank" rel="ugc" href="https://civitai.com/models/38290/belle-delphine">here</a>.) This LoRA is quite flexible, but this should be mostly thanks to SDXL, not really my specific training.As trigger word "Belle Delphine" is used. Additionally, “braces” has been tagged a few times in the source images, so you might also be able to use this as a tag during generation.I suggest a default LoRA weight of 1.I tested this LoRA with the base model + refiner, but I wasn’t really happy with that, as the refiner removed characteristics again.Most images were generated with <a target="_blank" rel="ugc" href="https://civitai.com/models/112902/dreamshaper-xl10">DreamShaper XL1.0 Alpha2</a>.Additionally tested were <a target="_blank" rel="ugc" href="https://civitai.com/models/114238/tdg8uus-sdxl10-mix">Tdg8uU's SDXL1.0 mix</a> and <a rel="ugc" href="https://civitai.com/models/117259/anime-art-diffusion-xl">Anime Art Diffusion XL</a> (the last 4 wide images are 4 different models). These images have been rendered using <a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI">ComfyUI</a> using quite a high-fidelity workflow (also depending on the image). But you can assume results were produced with a 1024x1024 base resolution (with different aspect ratio), then latent based img2img upscale with a factor of 1.5 and then also using ComfyUI-Impact-Pack for face details (if needed). The workflow should be embedded with the example images. It should work at &gt;= 8GB VRAM. To make use of it you will need the following custom ComfyUI nodes:<ul><li><a target="_blank" rel="ugc" href="https://github.com/ltdrdata/ComfyUI-Impact-Pack">ComfyUI-Impact-Pack</a></li><li><a target="_blank" rel="ugc" href="https://github.com/Ser-Hilary/SDXL_sizing">Automatic Sizing for SDXL</a></li><li><a target="_blank" rel="ugc" href="https://github.com/WASasquatch/was-node-suite-comfyui">WAS Node Suite</a> (only really used to save images with a timestamp, besides that not used)</li></ul> <h3 id="heading-96">Training:</h3>I assume some people might be interested in how this was trained, but I will have to disappoint you, as it was my first try with default settings and I am sufficiently happy with it, so no secret tips. In detail I used the following <a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss/blob/278b6937a09f5133d4b2a6b19145ad7c43135d49/presets/lora/SDXL%20-%20LoRA%20adafactor%20v1.0.json">base configuration file</a>.With those settings I used 151 uncropped, manually captioned high resolution images with bucketing in the resolution range 512-2048, 32 as bucket steps. The image dataset was set for 25 repetitions, additionally there was a regularization dataset (freshly generated with SDXL) with about 3000 images, set to 2 repetitions. Originally, I set training to 30 epochs (as that was the default of the previous mentioned config file), but around 8 epochs the results began to overbake. For that reason I stopped after 18 epochs (around 27000 steps). The final LoRA is then a merge of epoch 3 (10%), 6 (70%), 9 (12%) and 18 (8%). After that I resized it with a factor of 0.96, target rank 32 resulting in the ~40MB file. As you can see in the percentages (of the epochs in the merge), good LoRA results were already achieved relatively early.Training was done on a rented 3090 (cloud). Should you have any additional questions, feel free to ask.

Files with SHA256: 893907d225c6c1c2e70297bb7fbf88570b5b77be058d5b23d051f86f085c1ddd