CivArchive
    Mature Female - v2.0
    NSFW
    Preview 97990141
    Preview 97990230
    Preview 97990263
    Preview 97990335
    Preview 97990434
    Preview 97990485

    Man, I am so tired of models not being able to express age ranges between 15 and 65. The only settings they seem to know are "10yo", "15yo", "15yo with enormous honkers", and "grandmother who is basically a raisin."

    This LoRA tries to give you the ability to make women who actually look like adults. Without modification, Illustrious seems to think "mature female" just means "exaggerated proportions," so I tried to focus on mature facial features, not just colossal dobonhonkeros, because contrary to what anime might tell you, not all adult women have triple-F cups.

    VERSION 3 INFO:

    The keyword is "mtrfem". This is a new thing I'm trying where the concept is trained on an entirely new token instead of modifying the existing "mature female" term. The problem with "mature female" is that it has a lot of other biases and connotations other than age, so it tends to change characters a lot more. Hopefully this helps with that. You should be able to just use "1girl, mtrfem" by itself to easily just adult-ify any character with very little side effects.

    You can still use all the other normal modifiers as usual- curvy, plump, pubic hair, large breasts, etc- or leave them off entirely. There's also a secondary keyword, "mtrolder", which was intended to be for women who look over 50, but it doesn't quite reliably work. It can make the character look a bit less shiny and makeup-y, and give them a bit more of a worn, rougher appearance, which might work for certain styles. Feel free to experiment with it.

    If you don't mind adding some of the extra bias, combining "mtrfem" with "mature female" gives you an even stronger effect in just the right direction, in my opinion.

    As with version 1/2, you can push the strength above 1 if you just want to brute-force your way past any issues. I've tried it with strength up to 2 and it still gives me decently coherent images.

    The specifications for V3 are listed below. As always, please feel free to message me with any questions about the model or the training process.

    {
      "LoRA_type": "Standard",
      "LyCORIS_preset": "full",
      "adaptive_noise_scale": 0,
      "additional_parameters": "--log_prefix=mtrfem",
      "async_upload": false,
      "block_alphas": "",
      "block_dims": "",
      "block_lr_zero_threshold": "",
      "bucket_no_upscale": false,
      "bucket_reso_steps": 32,
      "bypass_mode": false,
      "cache_latents": true,
      "cache_latents_to_disk": true,
      "caption_dropout_every_n_epochs": 0,
      "caption_dropout_rate": 0,
      "caption_extension": ".txt",
      "clip_skip": 2,
      "color_aug": false,
      "constrain": 0,
      "conv_alpha": 4,
      "conv_block_alphas": "",
      "conv_block_dims": "",
      "conv_dim": 4,
      "dataset_config": "",
      "debiased_estimation_loss": false,
      "decompose_both": false,
      "dim_from_weights": false,
      "dora_wd": false,
      "down_lr_weight": "",
      "dynamo_backend": "no",
      "dynamo_mode": "default",
      "dynamo_use_dynamic": false,
      "dynamo_use_fullgraph": false,
      "enable_bucket": true,
      "epoch": 0,
      "extra_accelerate_launch_args": "",
      "factor": -1,
      "flip_aug": true,
      "fp8_base": false,
      "full_bf16": false,
      "full_fp16": false,
      "gpu_ids": "",
      "gradient_accumulation_steps": 1,
      "gradient_checkpointing": true,
      "huber_c": 0.1,
      "huber_schedule": "snr",
      "huggingface_path_in_repo": "",
      "huggingface_repo_id": "",
      "huggingface_repo_type": "",
      "huggingface_repo_visibility": "",
      "huggingface_token": "",
      "ip_noise_gamma": 0,
      "ip_noise_gamma_random_strength": false,
      "keep_tokens": 1,
      "learning_rate": 0.0001,
      "log_tracker_config": "",
      "log_tracker_name": "",
      "log_with": "tensorboard",
      "logging_dir": "D:/Coding/kohya_ss/logs",
      "loss_type": "huber",
      "lr_scheduler": "cosine",
      "lr_scheduler_args": "",
      "lr_scheduler_num_cycles": 1,
      "lr_scheduler_power": 1,
      "lr_warmup": 0,
      "main_process_port": 0,
      "masked_loss": false,
      "max_bucket_reso": 2048,
      "max_data_loader_n_workers": 0,
      "max_grad_norm": 1,
      "max_resolution": "1024,1024",
      "max_timestep": 1000,
      "max_token_length": 150,
      "max_train_epochs": 12,
      "max_train_steps": 0,
      "mem_eff_attn": false,
      "metadata_author": "restlessboy",
      "metadata_description": "",
      "metadata_license": "",
      "metadata_tags": "mtrfem, mtr_middleage, curvy, large breasts, plump, pubic hair, nude",
      "metadata_title": "mature female",
      "mid_lr_weight": "",
      "min_bucket_reso": 64,
      "min_snr_gamma": 0,
      "min_timestep": 0,
      "mixed_precision": "fp16",
      "model_list": "custom",
      "module_dropout": 0.1,
      "multi_gpu": false,
      "multires_noise_discount": 0,
      "multires_noise_iterations": 0,
      "network_alpha": 16,
      "network_dim": 16,
      "network_dropout": 0.1,
      "network_weights": "",
      "noise_offset": 0.0357,
      "noise_offset_random_strength": false,
      "noise_offset_type": "Original",
      "num_cpu_threads_per_process": 2,
      "num_machines": 1,
      "num_processes": 1,
      "optimizer": "AdamW",
      "optimizer_args": "",
      "output_dir": "D:/Coding/kohya_ss/outputs/model",
      "output_name": "mtrfem",
      "persistent_data_loader_workers": false,
      "pretrained_model_name_or_path": "D:/ai-models/Illustrious-v1",
      "prior_loss_weight": 1,
      "random_crop": false,
      "rank_dropout": 0.1,
      "rank_dropout_scale": false,
      "reg_data_dir": "",
      "rescaled": false,
      "resume": "",
      "resume_from_huggingface": "",
      "sample_every_n_epochs": 0,
      "sample_every_n_steps": 200,
      "sample_prompts": "masterpiece, best quality, 1girl, mtrfem --w 1024 --h 1024 --d 70 --l 6.0 --s 20 --n worst quality, lowres, 3d\nmasterpiece, best quality, 1girl, mtrfem, nude --w 1024 --h 1024 --d 71 --l 6.0 --s 20 --n worst quality, lowres, 3d\nmasterpiece, best quality, 1girl, mtrfem, bikini, pubic hair peek --w 1024 --h 1024 --d 72 --l 6.0 --s 20 --n worst quality, lowres, 3d",
      "sample_sampler": "euler_a",
      "save_every_n_epochs": 0,
      "save_every_n_steps": 0,
      "save_last_n_steps": 0,
      "save_last_n_steps_state": 0,
      "save_model_as": "safetensors",
      "save_precision": "fp16",
      "save_state": false,
      "save_state_on_train_end": false,
      "save_state_to_huggingface": false,
      "scale_v_pred_loss_like_noise_pred": false,
      "scale_weight_norms": 1,
      "sdxl": true,
      "sdxl_cache_text_encoder_outputs": false,
      "sdxl_no_half_vae": true,
      "seed": 7,
      "shuffle_caption": true,
      "stop_text_encoder_training_pct": 0,
      "text_encoder_lr": null,
      "train_batch_size": 2,
      "train_data_dir": "D:/Coding/kohya_ss/dataset/img",
      "train_norm": false,
      "train_on_input": false,
      "training_comment": "trigger: mtrfem",
      "unet_lr": null,
      "unit": 1,
      "up_lr_weight": "",
      "use_cp": false,
      "use_scalar": false,
      "use_tucker": false,
      "v2": false,
      "v_parameterization": false,
      "v_pred_like_loss": 0,
      "vae": "",
      "vae_batch_size": 0,
      "wandb_api_key": "",
      "wandb_run_name": "",
      "weighted_captions": false,
      "xformers": "xformers"
    }

    Description

    I do all my tagging manually, so it takes a while, but this is now updated with a larger 288-image dataset and better tagging. It's a little more resilient to being overridden by the general bias of checkpoints/other loras towards younger girls. I've included recommended tags that you might want to try with the "mature female" tag. Enjoy!

    FAQ

    Comments (4)

    MonkeyskillSep 4, 2025· 1 reaction
    CivitAI

    Amazing Job! Milfs for the Milf god!

    yemesavaSep 10, 2025· 1 reaction
    CivitAI

    Amazing!!

    sorzaSep 21, 2025
    CivitAI

    this works quite well But the fact that you trained it on the very word that makes girls look like exaggerated proportioned troglodytes makes it harder to use. please consider training it on something not directly on top of mature female. is there a way to change the activation word after training?

    restlessboy
    Author
    Sep 27, 2025· 1 reaction

    I don't believe there's any way to change the activation word after training, unfortunately. It's an interesting idea to use a new word that doesn't import the exaggerated features of the existing tag. I'm probably going to be uploading a new version of this over the next few days, so I may try a new word and see how it works. Check back soon.

    LORA
    Illustrious

    Details

    Downloads
    570
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/2/2025
    Updated
    6/11/2026
    Deleted
    -