THIS MODEL IS OLD/OUTDATED - you will have better results using more modern futa LoRAs.

This LoRA turns all girls into ???? girls. It was trained on Anything-V3.0 and can be used with other anime style models with ease.

Futa frontal 2 (LoRA) is the successor to Futa frontal 1 (hypernetwork), which moves away from 2D anime style to a more 2.5D anime style with more realism. It's also a 9MB LoRA instead of a 83MB hypernetwork, so that's nice.

Use instructions:

Include "<lora:futaFrontal2Lora_futaFrontal20:0.8>" anywhere in your prompt.

For best results use with models that focus on anime and use Danbooru tags. Good models that I've found to use it with are: Anything-V3.0 / 4.5 (4.5 for 2.5D, 3.0 for 2D), AbyssOrangeMix2, SimpMaker 3k3, kotosmix, and a mix of Anything + HassanBlend.

It performs poorly with models that have a photorealistic focus, such as: f222, SXD, HassanBlend, URPM, SD1.5. It can give women penises, but the penises look like lumps of flesh most of the time.

Recommended LoRA strength: 0.5-1.3 (start at 0.6, then increase the strength if there aren't enough penises, or to overcome any overtrained embeddings/models you are using)

Recommended Clip skip: 2 (1 works fine, but 2 seems to produce higher quality penises)

The training data was 512x??0 resolution so expect the best results when using that resolution, but it still works just fine at 512x512.

For prompting you need to include Danbooru tags since the LoRA was trained with WD14/Danbooru captions. At minimum you need to prompt for "futanari" or "penis", and the more specific you get with your prompt the better the results will be.

Embeddings that were created with photorealistic people also work well as long as they aren't overtrained.

You may need to use a VAE in order to not get desaturated images.

See below for more example prompts.

Training info:

The LoRA was trained on images captioned with WD14/Danbooru tags at 512x??0 resolution. In order to reduce the horror that is multiple penises growing out in multiple directions, the types of images were restricted to mostly frontal images. This helps make the ????? generation more consistent, but reduces the flexibility in poses you can ask it to do. Very large penises/boobs were filtered out to give the generated images more realistic sized anatomy.

Install:

For Automatic1111, install like any other LoRA by putting the .safetensor file in the stable-diffusion-webui/models/Lora folder.

Example prompts:

Many of the preview images used these prompts. For exact generation details, click the white ( i ) button on the bottom right of each preview image. All images were made with Clip skip 2, 512x??0 resolution, and upscaled by 2x with hires fix turned on with the 4x-UltraMix_Balanced upscaler.

Anything-V3.0 (preview pictures 1-8):

• <lora:Futa_frontal_2.0:0.6> 1girl, cute thin futanari, (huge long ?????:1.2), (waving:1.3), ????????, ?????????, grinning, head tilted, sitting on a wooden park bench, crop top, exposed navel, (legs spread:1.2), collarbone, ponytail, close-up ?????, blurry background, relaxed, sunlight, sun rays, 8k, ultra detail, masterpiece

  <lora:Futa_frontal_2.0:1> 1girl, (?????:1.2), futanari, futanari ????????????, on a bed in a bedroom, sunlight, soft lighting, blushing, grin, 8k, high detail, chiaroscuro

  <lora:Futa_frontal_2.0:1> 1girl, ?????, futanari, bikini, bikini aside, beach, ocean, sky, sunlight, 8k, high detail, realistic

  <lora:Futa_frontal_2.0:1> 1girl, ?????, futanari, Asuna, sitting on a bed in a bedroom, legs spread, wearing a blouse, covered ???????, 8k, high detail

  <lora:Futa_frontal_2.0:1> 1girl, ?????, futanari, (flaccid ?????:1.2), ????????, outdoors, lifting up skirt, view from below, embarrassed expression, 8k, high detail

  <lora:Futa_frontal_2.0:0.6> 1girl, ?????, futanari, ????, ???, ejaculation, futanari ????????????, breasts, ???????, beach, ocean, sky, sunlight, 8k, high detail, realistic

Anything-V4.5 (preview pictures 9-10):

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, wearing a mini skirt, skirt lift, ???????, pubic hair, view from below, city, embarrassed expression, blushing, sunlight, 8k, ultra detail

AbyssOrangeMix2_hard (preview pictures 11-16):

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, with cat ears, ?????????, collar with bell, arms up, playful, pubic hair, ????, 8k, ultra detail

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, showering, standing in a shower, arms behind head, (shower head:0.6), looking at viewer, wet skin, reflective skin, wet hair, ????, 8k, ultra detail

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, corset, latex, dominatrix, gothic architecture, rose, cape, precum drip, ???????, looking at viewer, squatting, legs spread, view from below, high heels, 8k, high detail

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, futuristic sexy revealing cyberpunk armor, high tech, circuit, neon lines, neon lights, foggy, signs, cyberpunk city in the background, 8k, ultra detail

Kotosmix (preview pictures 17-20):

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, futanari ????????????, wearing a miniskirt and thigh highs and lingerie, in a bedroom, (??? drip:1.2), ???, 8k, ultra detail

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, kneeling on a bed in a bedroom, chiaroscuro, sexy, removing jacket, 8k, ultra detail

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, sitting on a bed in a bedroom, ????, (perineum:1.2), ??? pool, ??? dripping, ??? on breast, ???, grabbing own breast, ????, 8k, ultra detail

• <lora:Futa_frontal_2.0:0.8> 1girl, ?????, futanari, skinny, thin, ponytail, sitting at a gym, projectile ???, ???, ??? dripping, ??????, sports bra, leggings, exercise equipment, exposed navel, collarbone, exhausted, (sweat:1.2), red hair, 8k, ultra detail

Negative prompt: lowres, low res, text, error, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, text, username, blurry, robot eyes, out of focus, censored, out of frame, b&w, mask, selfie

Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 7, Size: 512x??0, Denoising strength: 0.25, Clip skip: 2, Hires upscale: 2, Hires steps: 10, Hires upscaler: 4x-UltraMix_Balanced