A model I originally finetuned for making my cartoon/anime loras - but I accidently ended up just making a model and kept iterating on it. Be careful -- the danbooru images I used tends to move it into NSFW territory. Use danbooru tags. All images straight from the model -- no lora, or inpainting. Use Art by __________ as first tag. Can do SFW or NSFW. List of Artist Styles for Dynamic Prompt can be found at https://files.catbox.moe/w413ru.txt Comparisons can be found at https://mega.nz/folder/gf1xlKBI#FNRxuVUAU-fjx_khegWYhA
Blah blah blah buy me a kofi https://ko-fi.com/digimdkofi
Description
More epoch. More artist.
FAQ
Comments (5)
Any other interesting things you do during training? This model is crazy good at following prompts im not sure I have seen a model follow prompts this well while also being very creative and diverse with its outputs.
Sorry - don't think I did anything special. Loads of well-captioned data and training text encoder 2 at a real low rate. {
"__version": 1,
"training_method": "FINE_TUNE",
"model_type": "STABLE_DIFFUSION_XL_10_BASE",
"debug_mode": false,
"debug_dir": "debug",
"workspace_dir": "workspace\\run",
"cache_dir": "workspace-cache\\run",
"tensorboard": true,
"tensorboard_expose": false,
"continue_last_backup": true,
"base_model_name": "E:/ComfyUI_windows_portable/ComfyUI/models/checkpoints/AnimeConf3tti3.safetensors",
"weight_dtype": "BFLOAT_16",
"output_dtype": "FLOAT_32",
"output_model_format": "SAFETENSORS",
"output_model_destination": "E:\\OneTrainer\\models\\Finetune 2\\MoreBetterXL4.safetensors",
"gradient_checkpointing": true,
"concept_file_name": "training_concepts/concepts.json",
"concepts": null,
"circular_mask_generation": false,
"random_rotate_and_crop": false,
"aspect_ratio_bucketing": true,
"latent_caching": true,
"clear_cache_before_training": false,
"learning_rate_scheduler": "COSINE",
"learning_rate": 3e-05,
"learning_rate_warmup_steps": 200,
"learning_rate_cycles": 1,
"epochs": 250,
"batch_size": 4,
"gradient_accumulation_steps": 1,
"ema": "OFF",
"ema_decay": 0.999,
"ema_update_step_interval": 5,
"train_device": "cuda",
"temp_device": "cpu",
"train_dtype": "BFLOAT_16",
"fallback_train_dtype": "BFLOAT_16",
"only_cache": false,
"resolution": "896,960,1024,1088,1152",
"attention_mechanism": "SDP",
"align_prop": false,
"align_prop_probability": 0.1,
"align_prop_loss": "AESTHETIC",
"align_prop_weight": 0.01,
"align_prop_steps": 20,
"align_prop_truncate_steps": 0.5,
"align_prop_cfg_scale": 7.0,
"mse_strength": 1.0,
"mae_strength": 0.0,
"vb_loss_strength": 1.0,
"loss_scaler": "NONE",
"learning_rate_scaler": "NONE",
"train_unet": true,
"train_unet_epochs": 100000,
"unet_learning_rate": 3e-05,
"offset_noise_weight": 0.001,
"perturbation_noise_weight": 0.001,
"rescale_noise_scheduler_to_zero_terminal_snr": false,
"force_v_prediction": false,
"force_epsilon_prediction": false,
"min_noising_strength": 0.0,
"max_noising_strength": 1.0,
"noising_weight": 0.0,
"noising_bias": 0.5,
"unet_weight_dtype": "NONE",
"train_prior": true,
"train_prior_epochs": 10000,
"prior_learning_rate": null,
"prior_weight_dtype": "NONE",
"train_text_encoder": false,
"train_text_encoder_epochs": 30,
"text_encoder_learning_rate": null,
"text_encoder_layer_skip": 0,
"text_encoder_weight_dtype": "NONE",
"train_text_encoder_2": true,
"train_text_encoder_2_epochs": 100,
"text_encoder_2_learning_rate": 2e-07,
"text_encoder_2_layer_skip": 0,
"text_encoder_2_weight_dtype": "NONE",
"vae_model_name": "",
"vae_weight_dtype": "FLOAT_32",
"effnet_encoder_model_name": "",
"effnet_encoder_weight_dtype": "NONE",
"decoder_model_name": "",
"decoder_weight_dtype": "NONE",
"decoder_text_encoder_weight_dtype": "NONE",
"decoder_vqgan_weight_dtype": "NONE",
"masked_training": false,
"unmasked_probability": 0.1,
"unmasked_weight": 0.1,
"normalize_masked_area_loss": false,
"embedding_model_names": [],
"token_count": 1,
"initial_embedding_text": "*",
"embedding_weight_dtype": "FLOAT_32",
"lora_model_name": "",
"lora_rank": 16,
"lora_alpha": 1.0,
"lora_weight_dtype": "FLOAT_32",
"optimizer": {
"__version": 0,
"optimizer": "ADAMW",
"adam_w_mode": false,
"alpha": null,
"amsgrad": false,
"beta1": 0.9,
"beta2": 0.999,
"beta3": null,
"bias_correction": false,
"block_wise": true,
"capturable": false,
"centered": false,
"clip_threshold": 1.0,
"d0": 1e-06,
"d_coef": 1.0,
"dampening": null,
"decay_rate": -0.8,
"decouple": true,
"differentiable": false,
"eps": 1e-08,
"eps2": 0.001,
"foreach": false,
"fsdp_in_use": false,
"fused": true,
"growth_rate": "inf",
"initial_accumulator_value": null,
"is_paged": true,
"log_every": 0,
"lr_decay": null,
"max_unorm": null,
"maximize": false,
"min_8bit_size": 4096,
"momentum": null,
"nesterov": false,
"no_prox": false,
"optim_bits": 32,
"percentile_clipping": 100.0,
"relative_step": true,
"safeguard_warmup": true,
"scale_parameter": true,
"use_bias_correction": true,
"use_triton": false,
"warmup_init": false,
"weight_decay": 0.01
},
"optimizer_defaults": {},
"sample_definition_file_name": "training_samples/samples.json",
"samples": null,
"sample_after": 5,
"sample_after_unit": "EPOCH",
"sample_image_format": "PNG",
"samples_to_tensorboard": true,
"non_ema_sampling": true,
"backup_after": 30,
"backup_after_unit": "MINUTE",
"rolling_backup": true,
"rolling_backup_count": 1,
"backup_before_save": false,
"save_after": 50,
"save_after_unit": "NEVER"
} training settings.
@spadira272 Dude you rock Thank you!
wow this model is amazing. how did you make it do perfect upside down faces. and not just faces the entire body upside down comes out perfect
Very good model, have you thought of doing something like HLL for pony?
Details
Files
AnimeConf3tti3.yaml
Mirrors
ft65p.yaml
LifeIs2uffering.yaml
ishikeistyle.yaml
confetti4comrades.yaml
Battlemap_X2_1.yaml
Battlemap_4_400.yaml
homosimileAgainPony_v10.yaml
maleaiPonyIllustriousNoobai_v10.yaml
chloeBennet_v10.yaml
Male AI 1_1.yaml
2026_06_17 realistic.yaml
anim3conf3tticomrade.yaml
AnimeConf3tti3.yaml
Con5ti.yaml
homosimileAgainPony_v10.yaml
confetti4you.yaml
MoreConfettialpha.yaml
AnimeConfetti.yaml
2025_09_29 LongTermGoal 2.yaml
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.



















