Amber is a LoRA for Pony models that you can use to generate realistic images of beautiful women with pale skin.<h2 id="recommended-usage-xinxp37ab">Recommended usage</h2><ol><li>Put "score_9, score_8_up, score_7_up, realistic, ccamber, 1girl," at the beginning of the positive prompt.</li><li>Set LoRA weight to 0.8 and tweak from there.</li><li>Use highres fix.</li><li>If freckles are too much (especially when you use ComfyUI), specify the word "freckles" in negative prompt and adjust the weight of this word.</li><li>Use with CyberRealistic Pony v7 as the main model (<a target="_blank" rel="ugc" href="https://civarchive.com/models/443821/cyberrealistic-pony">https://civarchive.com/models/443821/cyberrealistic-pony</a>) as it was used for LoRA training.</li></ol><h2 id="example-prompt-khfzxyiya">Example prompt</h2>Positivescore_9, score_8_up, score_7_up, realistic, ccamber, 1girl,portrait,22yo woman, (beautiful, sweet),dress, mansion&lt;lora:ccamber_v2:0.8&gt;Negativescore_6, score_5, score_4, ugly<h2 id="how-this-lora-was-created-8djj1jsmx">How this LoRA was created</h2><h3 id="dataset-creation-nvhc6ypr1">Dataset creation</h3>Main dataset<ul><li>115 synthetic safe-for-work images of women were generated using the epiCRealismXL v7-Final Destination model (<a target="_blank" rel="ugc" href="https://civarchive.com/models/277058?modelVersionId=489217">https://civarchive.com/models/277058?modelVersionId=489217</a>) with the words "22yo woman" included in the prompt.</li><li>Faces were further tweaked using the The Realistic Vision V6.0 B1 (VAE) model (<a target="_blank" rel="ugc" href="https://civarchive.com/models/4201?modelVersionId=245598">https://civarchive.com/models/4201?modelVersionId=245598</a>) with the words "22yo woman" included in the prompt.</li></ul>Regularization dataset<ul><li>199 synthetic safe-for-work images of women were generated using the epiCRealismXL v7-Final Destination model (<a target="_blank" rel="ugc" href="https://civarchive.com/models/277058?modelVersionId=489217">https://civarchive.com/models/277058?modelVersionId=489217</a>) with the words "22yo woman" included in the prompt. Face enhancement was not done for this dataset.</li></ul><h3 id="training-j6ytn75vv">Training</h3>Kohya trainer was run with the following parameters:<pre><code>accelerate launch --num_cpu_threads_per_process=2 "./sdxl_train_network.py" \
 --enable_bucket \
 --min_bucket_reso=256 \
 --max_bucket_reso=2048 \
 --pretrained_model_name_or_path="&lt;path to model dir&gt;/cyberrealisticPony_v7.safetensors" \
 --dataset_config=&lt;path to dataset dir&gt;/ccamber_v2/ccamber_v2.toml \
 --resolution="1024,1024" \
 --output_dir=&lt;path to LoRA dir&gt; \
 --output_name=ccamber_v1 \
 --logging_dir="output/log" \
 --network_alpha=128 \
 --save_model_as=safetensors \
 --network_module=networks.lora \
 --text_encoder_lr=0.0001 \
 --unet_lr=0.0001 \
 --network_dim=128 \
 --lr_scheduler_num_cycles="10" \
 --no_half_vae --learning_rate="0.0001" \
 --lr_scheduler="cosine" \
 --train_batch_size="1" \
 --max_train_steps="5000" \
 --save_every_n_epochs="1" \
 --mixed_precision="bf16" \
 --save_precision="bf16" \
 --cache_latents \
 --cache_latents_to_disk \
 --optimizer_type="Adafactor" \
 --gradient_checkpointing \
 --optimizer_args scale_parameter=False relative_step=False warmup_init=False \
 --max_data_loader_n_workers="0" \
 --bucket_reso_steps=64 \
 --xformers \
 --bucket_no_upscale \
 --vae &lt;path to VAE&gt;</code></pre>

Amber

Cleaned up dataset to hopefully generate more consistent images.

Files with SHA256: d4891489adc51ac7a8201d651e5073728c506e5f5d6389084333dcf9b1f82651