Ver NOOB V-PRED V3 EQ
Trained with improved parameters based on NOOB V-PRED EQ VAE exprimental version: https://civarchive.com/models/1858821/noobai-v-pred-10-with-eq-vae
In this version, I subdivided the trigger words for more accurate generate. 10 triggerwords set. See the examples first.
NOOB V-PRED V2
Trained base on NOOB V-PRED 1.0 with Kohya SS. See about this version for training details. Dataset updated and triggerwords changed.
The main upgrade of V2 is the understanding of multi futa scenes after changing training parameters and dataset, as well as the compatibility of finetuned checkpoints and multi loras. But it still need some trials to get the desired generation.
NOOB V-PRED V1
Trained base on NOOB V-PRED 1.0 with OneTrainer. See about this version for training details. Dataset updated and triggerwords changed.
The reason for this update is when I applied multiple Loras at the same time, increasing the weight of gigantic penis to 2 still couldn't achieve the result I needed. I think this Lora will solve the problem of giant pp generation when multiple Loras work together. After testing, the effect is good.
About PonyXL V2.0
Used regularization techniques and updated dataset to get less influence on artstyle and better penis shape. Also improved the effect when using other loras together. I chose Lora with less bake to achieve better interaction, but the disadvantage is that in some scenarios, the weight of penis needs to be increased.
Added "horse penis" for penis type.
About PonyXL
Used PonyXL to train this one. Updated dataset to achieve better performance. It enhanced this concept on PonyXL and related models.
Adding a concept: gigantic_dark
About V3.0
Updated dataset and used multires technique in this version. It would bring more light and shadow with quality tag, such as masterpiece, highly_detailed, best_quality ,etc.
Bewared that may affect the result of generation. You may get chaos result if you use these tags and overweight them.
Multires technique parameter:
multires_noise_iterations="6"
multires_noise_discount=0.3
IMPORTANT: Now I set the trigger words as gigantic_penis and gigantic_penis_ejaculation.
About V2.0
In v2.0, I removed all the images with black bar censor and retrain this locon. That will do help to get better result. The annoying censor bar won't be there.
I set batch size = 2 to acclerate training. The final results are good.
Introduction
This LyCORIS is used to generate futanari with gigantic penis. If you put it in a long picuture like (1024X768), you can get a massive long cock. You may find the faces are blurry but you can fix it with inpaint or img to img. Adding hyper futanari LyCORIS can improve the texture of penis.
How to use
I recommand using it in weight 0.75-0.85. When working with other loras, it is not easy to control. But you can get what you want after some tries.
You may get good result in lower weight when use it alone. Stiil, it needs more tests.
Training details:
Traingset is of about 145 images. They are mirrored before training. Half of them are chosen as regularization images. Total steps are about 8800.
Regularization = true
resolution=512
batch_size=2
epoch=10
network_dim=32
network_alpha=32
clip_skip=2
Using AdamW8bits as optimizer:
lr="1e-4"
unet_lr="1e-4"
text_encoder_lr="1e-5"
Locon parameters:
conv_dim=4
conv_alpha=4
Description
{
"LoRA_type": "LyCORIS/LoCon",
"LyCORIS_preset": "full",
"adaptive_noise_scale": 0,
"additional_parameters": "--zero_terminal_snr",
"ae": "",
"apply_t5_attn_mask": false,
"async_upload": false,
"block_alphas": "",
"block_dims": "",
"block_lr_zero_threshold": "",
"blocks_to_swap": 0,
"bucket_no_upscale": true,
"bucket_reso_steps": 64,
"bypass_mode": false,
"cache_latents": false,
"cache_latents_to_disk": false,
"caption_dropout_every_n_epochs": 0,
"caption_dropout_rate": 0,
"caption_extension": ".txt",
"clip_g": "",
"clip_g_dropout_rate": 0,
"clip_l": "",
"clip_skip": 0,
"color_aug": false,
"constrain": 0,
"conv_alpha": 1,
"conv_block_alphas": "",
"conv_block_dims": "",
"conv_dim": 16,
"cpu_offload_checkpointing": false,
"dataset_config": "",
"debiased_estimation_loss": false,
"decompose_both": false,
"dim_from_weights": false,
"discrete_flow_shift": 3,
"dora_wd": false,
"double_blocks_to_swap": 0,
"down_lr_weight": "",
"dynamo_backend": "no",
"dynamo_mode": "default",
"dynamo_use_dynamic": false,
"dynamo_use_fullgraph": false,
"enable_all_linear": false,
"enable_bucket": true,
"epoch": 10,
"extra_accelerate_launch_args": "",
"factor": -1,
"flip_aug": false,
"flux1_cache_text_encoder_outputs": false,
"flux1_cache_text_encoder_outputs_to_disk": false,
"flux1_checkbox": false,
"fp8_base": false,
"fp8_base_unet": false,
"full_bf16": false,
"full_fp16": false,
"gpu_ids": "",
"gradient_accumulation_steps": 1,
"gradient_checkpointing": true,
"guidance_scale": 3.5,
"highvram": false,
"huber_c": 0.1,
"huber_scale": 1,
"huber_schedule": "snr",
"huggingface_path_in_repo": "",
"huggingface_repo_id": "",
"huggingface_repo_type": "",
"huggingface_repo_visibility": "",
"huggingface_token": "",
"img_attn_dim": "",
"img_mlp_dim": "",
"img_mod_dim": "",
"in_dims": "",
"ip_noise_gamma": 0.1,
"ip_noise_gamma_random_strength": false,
"keep_tokens": 1,
"learning_rate": 1,
"log_config": false,
"log_tracker_config": "",
"log_tracker_name": "",
"log_with": "tensorboard",
"logging_dir": "G:/kohya_ss/outputs",
"logit_mean": 0,
"logit_std": 1,
"loraplus_lr_ratio": 0,
"loraplus_text_encoder_lr_ratio": 0,
"loraplus_unet_lr_ratio": 0,
"loss_type": "l2",
"lowvram": false,
"lr_scheduler": "cosine",
"lr_scheduler_args": "",
"lr_scheduler_num_cycles": 1,
"lr_scheduler_power": 1,
"lr_scheduler_type": "",
"lr_warmup": 0,
"lr_warmup_steps": 0,
"main_process_port": 0,
"masked_loss": false,
"max_bucket_reso": 4096,
"max_data_loader_n_workers": 0,
"max_grad_norm": 1,
"max_resolution": "1024,1024",
"max_timestep": 1000,
"max_token_length": 225,
"max_train_epochs": 0,
"max_train_steps": 0,
"mem_eff_attn": false,
"mem_eff_save": false,
"metadata_author": "",
"metadata_description": "",
"metadata_license": "",
"metadata_tags": "",
"metadata_title": "",
"mid_lr_weight": "",
"min_bucket_reso": 1024,
"min_snr_gamma": 3,
"min_timestep": 0,
"mixed_precision": "bf16",
"mode_scale": 1.29,
"model_list": "custom",
"model_prediction_type": "sigma_scaled",
"module_dropout": 0,
"multi_gpu": false,
"multires_noise_discount": 0.1,
"multires_noise_iterations": 6,
"network_alpha": 1,
"network_dim": 32,
"network_dropout": 0.1,
"network_weights": "",
"noise_offset": 0,
"noise_offset_random_strength": false,
"noise_offset_type": "Original",
"num_cpu_threads_per_process": 2,
"num_machines": 1,
"num_processes": 1,
"optimizer": "Prodigy",
"optimizer_args": "\"weight_decay=0\" \"d_coef=1\" \"use_bias_correction=True\" \"safeguard_warmup=True\" \"betas=0.9,0.999\" \"decouple=True\"",
"output_dir": "G:/kohya_ss/outputs",
"output_name": "gigantic_futanari-NOOB-VPRED1.0-V1-locon-dim32alpha1-drop0.1-SNR3",
"persistent_data_loader_workers": false,
"pos_emb_random_crop_rate": 0,
"pretrained_model_name_or_path": "G:/stable-diffusion-webui-reForge/models/Stable-diffusion/NAI/noobaiXLNAIXL_vPred10Version.safetensors",
"prior_loss_weight": 1,
"random_crop": false,
"rank_dropout": 0,
"rank_dropout_scale": false,
"reg_data_dir": "",
"rescaled": false,
"resume": "",
"resume_from_huggingface": "",
"sample_every_n_epochs": 1,
"sample_every_n_steps": 0,
"sample_prompts": "hyperfonf, long hair, breasts, multiple girls, blonde hair, 2girls, nipples, purple hair, thighs, wings, barefoot, armor, huge breasts, uncensored, colored skin, chain, thick thighs, huge testicles, feathers, feathered wings, multiple penises, veins, futanari, angel wings, veiny penis, angel, battle, precum, duel, futa with futa, gold armor, angel and devil, hyper penis, hyper stomach bulge, hyper insertion, futa on futa\uff0c masterpiece, absurdres --n worst quality, low quality, old, lowres, bad anatomy, bad hands, multiple views, text, signature, watermark, logo, username, mammal, anthro, furry, ambiguous form, feral, semi-anthro, censored, mosaic censoring, bar censor --w 1216 --h 1024 --l 5 --s 28",
"sample_sampler": "euler_a",
"save_clip": false,
"save_every_n_epochs": 1,
"save_every_n_steps": 0,
"save_last_n_epochs": 0,
"save_last_n_epochs_state": 0,
"save_last_n_steps": 0,
"save_last_n_steps_state": 0,
"save_model_as": "safetensors",
"save_precision": "bf16",
"save_state": false,
"save_state_on_train_end": false,
"save_state_to_huggingface": false,
"save_t5xxl": false,
"scale_v_pred_loss_like_noise_pred": false,
"scale_weight_norms": 0,
"sd3_cache_text_encoder_outputs": false,
"sd3_cache_text_encoder_outputs_to_disk": false,
"sd3_checkbox": false,
"sd3_clip_l": "",
"sd3_clip_l_dropout_rate": 0,
"sd3_disable_mmap_load_safetensors": false,
"sd3_enable_scaled_pos_embed": false,
"sd3_fused_backward_pass": false,
"sd3_t5_dropout_rate": 0,
"sd3_t5xxl": "",
"sd3_text_encoder_batch_size": 1,
"sdxl": true,
"sdxl_cache_text_encoder_outputs": false,
"sdxl_no_half_vae": true,
"seed": 31337,
"shuffle_caption": true,
"single_blocks_to_swap": 0,
"single_dim": "",
"single_mod_dim": "",
"skip_cache_check": false,
"split_mode": false,
"split_qkv": false,
"stop_text_encoder_training": 0,
"t5xxl": "",
"t5xxl_device": "",
"t5xxl_dtype": "bf16",
"t5xxl_lr": 0,
"t5xxl_max_token_length": 512,
"text_encoder_lr": 1,
"timestep_sampling": "sigma",
"train_batch_size": 2,
"train_blocks": "all",
"train_data_dir": "F:/gigantic_futanari_V2",
"train_double_block_indices": "all",
"train_norm": false,
"train_on_input": true,
"train_single_block_indices": "all",
"train_t5xxl": false,
"training_comment": "",
"txt_attn_dim": "",
"txt_mlp_dim": "",
"txt_mod_dim": "",
"unet_lr": 1,
"unit": 1,
"up_lr_weight": "",
"use_cp": false,
"use_scalar": false,
"use_tucker": false,
"v2": false,
"v_parameterization": true,
"v_pred_like_loss": 0,
"vae": "",
"vae_batch_size": 0,
"wandb_api_key": "",
"wandb_run_name": "",
"weighted_captions": false,
"weighting_scheme": "logit_normal",
"xformers": "xformers"
}FAQ
Comments (35)
one thing, the V-Pred V2 is labelled as SD1.5 instead of Illustrious/NoobAi
Sorry, corrected now.
@DeepDark_Fantasy514 all good, just thought it was odd saying SD1.5, i was like 'did he go back or something' then i read the title and knew it was a misinput.
@DeepDark_Fantasy514 it's odd that the site doesn't auto-detect the usage of this lora/Lycon from images i upload, as i have to manually add the resource to each image, i don't know if it's a weird feature or just a bug on the site's end.
@NightKillers101 I also run into such problem. Just another Civitai bug I think.
@DeepDark_Fantasy514 right, thanks, i was wondering if it was just me, i hate the bug, hopefully it's patched sooner or later.
somehow when i remove the Gigantic_horse prompt the penis does not increase in size? Any idea
I don't know what you mean. Gigantic_horse is a triggerword used in the version from long time ago. It was used to change the type of penis when the checkpoints (SD1.5 and Pony) didn't understand the concept of horse penis. NOOB can understand that so I removed it in NOOB version. Whatever it does not directly affect the size. If you want to change the size, the best way is to change the weight of keyword or the lora.
@DeepDark_Fantasy514 So i need to add like huge penis and such and just add the for example (huge penis: 1.5) ?
@Uncoloredenoki I assume you are using PonyXL V2.0 version.
You can use gigantic_cock here and upweight it, and gigantic penis and huge penis you mentioned will also be useful.
@Uncoloredenoki Here, if you want the horse cock. Put the gigantic_horse and horse penis weight higher than "gigantic penis" or gigantic_cock, otherwise you may get a weird type penis between horse penis and humaniod penis.
@DeepDark_Fantasy514 @DeepDark_Fantasy514 So i am using the base model hassakuXLIllustrious_v21fix what lycoris of yours would you recommend to take with that?
@Uncoloredenoki Just NOOB V-PRED V2 would be fine. Check its examples and type in horse penis in the positive prompt if you need it. hassakuXLIllustrious understands the concept of horse penis.
@DeepDark_Fantasy514 Can you maybe explain me how to use every tag from the Version you said since i somehow dont get it to work for me, i dont know how to combine them and what some of them trigger
@Uncoloredenoki Trigger words goes as it says. Gigantic ones bigger than the hyper ones. Erect means erection and flaccid means flaccidity. I think the samples have spoken for themsevles. You may firstly describe what you want to generate or upload your trials to the review of this. So I may figure out what's the real problem and its solution. For instance, assuming using hassakuIL2.2 and Giganticfutanari NOOB V-PRED V2, if you need to generate a girl with hyper horse penis, just type in hypererect, horse penis, hyper penis in the positive prompt and generate.
@DeepDark_Fantasy514 i did just upload one, where the girl is a bunny girl with a horse penis, i cant get like massive sizes this is like the biggest i ever had gotten with this lora,planning to go like twice or 3 times this size
@Uncoloredenoki Just checked your image. You may need remove other penis related keywords, only keep gigantic penis, horse penis and giganticerect especially the penis in the negatice prompt. Excess triggerwords will only disrupt your generation. Add "from side" ,"from above" and "wide shot", Upweight the gigantic penis to 1.2 or higher. Then adjust your image ratio to 16:9 or 9:16. Check my post: https://civitai.com/posts/16042347
@DeepDark_Fantasy514 So i keep the penis tag in the negative prompts or do i remove it ?
@Uncoloredenoki remove it.
@DeepDark_Fantasy514 I uploaded now another one could you tell me if this is correct or not cause the size is still quite small and i copied the second one you send me
@Uncoloredenoki The problem is the ratio. You can see that your image is almost square. No room for the enlarged penis so the checkpoint won't do it. Lengthen your image using ratio like1344X768 or 768 X 1344.
@DeepDark_Fantasy514 Sorry for all this questions i started Stable diffusion 3 days ago im really new to all of this
@DeepDark_Fantasy514 Could i maybe add you on discord so its simpler with the sharing and questions ?
@Uncoloredenoki Sorry. I’m not on Discord much. My schedule's all over the place, so my responses can be super quick one day and then radio silence the next month. If you need generation tips though, just hit me up in the comments under this Lyco. I will reply as I see it.
@DeepDark_Fantasy514 how do i make it work with the Succubus tail job you did with the one, i get the tail and such there but its does not devour the penis, i send in what i mean, any ideas ?
@Uncoloredenoki A simple but somewhat counterintuitive idea. Refer to the example image, do not include words related to 'penis' in the positive prompt, instead using hyerpenetration and hyper insertion. The reason is: penis in trained lyco includes all parts, from shaft to glan. When applying lyco checkpoints will draw all the parts. That's also why I set the trigger word hyperpenetration separately. Checkpoints are not smart enough to understand the concept of hyper penis wrapped in tail after Lora training. So you need to use the separately trained hyperpenetration concept to make it draw wrapped penis like objects.
In large-scale fine-tuning of checkpoints, this problem does not occur because they have a sufficient understanding of the concepts related. However, in small-scale fine-tuning models like lora or lyco, it is difficult for newly added concepts to interact with previous concepts.
@DeepDark_Fantasy514 I dont know how checkpoints work, can you explain me how to use them and how they are used
@Uncoloredenoki Wow, that would be a long story. Perhaps we need to go back to the basic principles of stable diffusion to understand 'how the checkpoints are used'. I don't think you need such a basic principle explanation, and I haven't learned so deeply to explain it clearly. So next, I will share some simplified content that I am aware of, but I cannot guarantee its accuracy. My understanding is that checkpoint is a painter who has undergone extensive training. He has basic techniques and some concepts that he already knows, but some concepts he do not understand, so he needs us to provide certain hints (lora) to help him learn. Unfortunately, some of our hints were too difficult for him to establish a connection between Lora and other concepts, so he said, 'Okay, I'll draw as you say. If you want a super big penis, I'll draw a super big penis, and if you want a tail, I'll draw a tail. But I don't care what happens between the tail and penis. ” So we will have a hyper penis that cannot be wrapped up. At this point, we can only change our approach and tell him, 'Hey bro, what you're actually drawing is a belly that's been roughly inserted by a super big penis, or something similar.” So he will find a way to draw this concept, approaching tailjob, so we can get a big tailjob.
@DeepDark_Fantasy514 so how do i make it work, like where do i input these checkpoints
@Uncoloredenoki See this tutorial:https://www.youtube.com/watch?v=zqgKj9yexMY&list=PL-pohOSaL8P_VxpGxcay1EJFtqX4m8WqZ
could you upload it to tensor art please?
You get it:https://tensor.art/models/856634667724470343
@DeepDark_Fantasy514 many thanks man
plz one for illustrious :c
NOOB lora works fine with most illu realted checkpoints expect Miaomiao.