Realistic - God 2

Join my new Discord here! https://discord.gg/cnVvnkgtPb

@mods - these images are trained from public images from online sources (ex. r/). Does not violate any ToS. Pls do not <3

God 2 - an iconic . This is truly incredible. One of the most deserving to be labeled a God . (All images & models are prompted with descriptions & trained with models!)

Read Me!

DISCLAIMER: IF YOU DON'T INPAINT, YOUR RESULTS MIGHT !

TWO VERSIONS (I RECOMMEND USING BOTH):

• GodPussy2 v1: Trained on SD 1.5 base model, compatible with all models, still works better on photorealistic models

  • USE THIS LORA FOR INITIAL TXT2IMG GENS

  • For txt2img & img2img: LoRA weight between 0.2-0.8 for best results

    • v1 is meant to be strong, which lets you mix with other LoRAs (ex. face) much more easily

• GodPussy2 v1 Inpainting: Trained on LazyMix+, the best all around checkpoint for highly-detailed

  • USE THIS LORA FOR INPAINTING AND/OR UPSCALING

  • For txt2img & img2img: LoRA weight between 0.8-1 for best results (higher values can affect the skin texture more)

  • Note: still uses LazyMix+ v2.0, I tried v3.0b but I didn't like it as much

Version 1

• GodPussy2 v1

  • SD 1.5 base model (best compatibility across all checkpoint models - still works better on photorealistic models)

  • 576x576

  • 1 epoch, 100 repeats, 33 images

    bucket 0: resolution (448, 576), count: 500

    bucket 1: resolution (448, 704), count: 1300

    bucket 2: resolution (512, 640), count: 200

    bucket 3: resolution (576, 576), count: 1200

    bucket 4: resolution (640, 448), count: 100

• GodPussy2 v1 Inpainting

  • 512x768 (MUST USE 512x768 2:3 ASPECT RATIO FOR BEST RESULTS!)

  • 5 epochs, 25 repeats, 33 images

• Read Recommended Settings and Notes for tips on how to get the best results

Recommended Settings

• NEW: for some reason, 'rear ' doesn't work very well with v1. Works better on v1 Inpainting

• ChilloutMix derived models (ex. MajicMix, PerfectWorld, RevAnimated) occasionally struggle with this LoRA for who knows why, be patient and INPAINT

• Make sure you use 'godpussy2' in positive, and you can also use trigger word 'open_' in positive or negative to finetune result

• For txt2img: v1 LoRA weight of <=0.8

  • You can generate the composition in txt2img, then edit further with img2img/inpainting with this LoRA or v1 Inpainting LoRA

  • HiRes Upscale: Denoising <=0.3 (higher than 0.3 can result in double - not what you want). I usually use 0.2

  • Put (fingers:1.5) in negative prompt if you don't wanna deal with stupid fingers

  • For animated models, you may need to raise the weight even more depending on the composition (WARNING: CAREFUL NOT TO GO TOO HIGH, MAY NEGATIVELY IMPACT IMAGE QUALITY), then inpaint for a more perfect result

• For img2img: v1 LoRA weight of 0.4-0.8

• Inpainting: use GodPussy2 v1 Inpainting, LoRA weight of 0.8-1

  • Masked Content = Original

  • Inpaint area: Whole Picture for better integration, Only Masked for just masked area - may want to adjust prompt if using Only Masked

  • Padding medium size (ex. 80-100), denoising around 0.2-0.8 depending on how drastic of a change you want (0.45 is a good number)

• Best models: Real Hot Mix, LazyMix+

  • Real Hot Mix is a mix I created to include LazyMix with some other photoreal models

  • LazyMix+ utilizes Subreddit v3, where you can add "A photo from the ____ subreddit, ____ quality" to the prompt (ex. , GodPussy, or AsiansGoneWild) for more variation

Notes

• Tips on getting extremely high-quality results:

  1. Generate large batch in txt2img, set v1 LoRA weight lower for full body pics (somewhere between 0.2-0.6). I use Real Hot Mix for all txt2img gens

     • HIGHLY recommend using ControlNet, I mostly use softedge & openpose for composition, recently started using t2ia color grid as well

  2. Select best pics, crop/edit outside of SD, inpaint as needed (ex. v1 Inpainting LoRA)

  3. Insert edited pic into img2img with same/mostly same prompt, LoRA weight around 0.4-0.6

  4. Denoising between 0.2-0.5, generate

  5. Select best pics, crop/edit outside of SD, inpaint as needed (ex. v1 Inpainting LoRA)

  6. Repeat steps 3-5 as many times as needed

  7. Insert edited pic into img2img with same/mostly same prompt

  8. Denoising between 0.1-0.3, upscale: (1.25x-2x), generate

• Don't forget to go outside and touch grass

Background

I'm a dude and I've seen a lot of . I believe they're all unique, like a face. But different also have very repeatable looks/themes. The plan is to create a set of LoRAs with categories of types using images of the same model. Eventually it will be like a catalog to select which LoRA for the type of you're looking for.

Contact me at [email protected]