Everlasting Summer Olga by GraffMetal (non-commercial use only)

Warning: This model is created for non-commercial use only! By downloading and using this model, you agree to these terms. All images created by me are also copyrighted. For commercial use of my model, images, or generated images, or for selling my images or generated images, I will ban your public accounts (TG channel, Patreon, etc.). Please respect the intellectual property of the original creator/developer of this character! I created this LoRa to generate good art, sprites, and to help mod creators, not for business purposes.

P.S. For using my models or images outside of CivitAI, please provide a link to my model, my CivitAI profile, or tag me as the author. About images, I don't like porn and ugly things, I'm prefer sfw, questionable, aesthetic and ero-aesthetic images. If you like this model, BUZZ IT :)

Olga from visual novel Everlasting Summer

V2.4

Necessarily use with my Art Style LoRa and model Prefect Pony XL V2 cleaned style !!!

Lora trained on 52 images. Reduce dataset, but works better for my opinion, use my settings for correct work.

I recommend using Hires.fix and Adetailer extensions for best results (but I use only adetailer)

Trigger Words:

1. olga, long hair, brown hair, green eyes, bangs, parted bangs, straight hair, tan skin, dark skin,

2. olga, long hair, brown hair, green eyes, bangs, parted bangs, straight hair, tan skin, dark skin, solo, medium breasts,

   +mature girl,

   Negative prompt- jewelry, earrings,

   ! Optional words

   mature girl - with this works better, not included in the dataset

   dark skin - optional, but canonical and the model works better with it, included in the dataset

   tan skin - optional, but works much better with it, in my opinion, word is canonical, while Semene is carrying sugar, Olga is chill on the beach :D if Im not mistaken, it is included in the dataset

   large breasts - included in the dataset

   medium breasts - included in the dataset, less in dataset than large, but on my opinion this looks more aesthetic

Settings

Olga LoRa strength 0.7-1, my Art Style LoRa strength 0.7-1, I prefer Olga 0.7 and Art style 1 or 0.7 and 0.7 ,or 1 and 1 ,or 1 and 0.7

Sampling method + Schedule type + Sampling steps

1.DPM++ 2M and Align Your Steps, Sampling steps 28

2.DPM++ 2M and Karras, Sampling steps 30

3.Euler a and Automatic, Sampling steps 40-50

CFG Scale 7

Clip skip 2

Hires.fix: Hires steps 20 Denoising strength 0.4 Upscaler 4x-Ultrasharp or 4x NMKD Superscale, also ESRGAN_4x are not bad

I highly recommend watching my metadata on image!

V1

Lora trained on 70 images
I recommend using Hires.fix and Adetailer extensions for best results(Adetailer face_yolov8s/n works not good, use mediapipe_face_mesh_eyes_only or Eyes detection (Adetailer))
Trigger Words: olga, green eyes, brown hair, long hair, 1girl, solo,
also you can use mature trigger word for more adult looking but figure be bigger, for better figure looking use mature, fit body, skinny, medium breasts, also you can add in negative prompt big breasts

I use trigger words:

1. olga, green eyes, brown hair, long hair, 1girl, solo, lips, (some times wavy hair, with this hair she looks gorgeous)

2. olga, green eyes, brown hair, long hair, 1girl, solo, lips, mature, fit body, skinny, medium breasts, in negative prompt big breasts

3. olga, brown hair, long hair, green eyes, bangs, parted bangs, lips, parted lips,

4. olga, brown hair, long hair, green eyes, bangs, parted bangs,

5. olga, brown hair, long hair, green eyes, bangs,

6. olga, brown hair, long hair, green eyes,

Hires.fix: Hires steps 20 Denoising strength 0.4 Upscaler 4x-Ultrasharp or 4x NMKD Superscale, also ESRGAN_4x are not bad

Lora strength 0.5 - 0.7 (I prefer 0.7) on 0.8 can be body artifact

Clip skip 2