DragonBall girls. You can change face with embedding on adetailer face. All images are generated with A1111 on PC with my checkpoint. Work better with simple negative prompt (3 embeddings) and detailed positive prompt.
If you like the character, always leave a like and a picture, and if you give some buzz you will allow me to train other LORAs or MODELS. Downloading only is not useful. For model visibility, remember to rate the images in my gallery and leave a like. Thank you.
If you like my models, please read, it's important. | Civitai
IMPORTANT : If you can, please donate me some buzz to allow me to train other loras and generate other content. Please always vote with a like for the model and images (not only download), it costs nothing and help me continue to share. Thank you.
IMPORTANT: All my models work only if the trigger words and the recommended negative and positive prompts and embeddings are used. If you have a different result from the images, it is because you have not used the recommended settings. Too many loras and/or embeddings can ruin the model's result. To have images like the ones I made, use ONLY the recommended settings. Thanks.
SD Negative prompt (for ALL): (CyberRealistic_Negative-neg), (worst quality:2), (low quality:2), (bad-hands-5, lowres, anime, bad anatomy, deformed, missing fingers, extra digit, fewer digits, cropped,watermark,username),(blurry:1.4) ,easynegative, unreal engine, (nipples), multiple hands
SD Positive prompt (VIDEL) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of videl, black hair, twintails, parted bangs, blue eyes, long eyelashes, black gym shorts, white shirt, short sleeves, gloves, medium breasts, slim body, hourglass body shape, big thighs,....
Weight 0.95 CFG 7.0
SD Positive prompt (MAI) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of mai, black hair, long straight hair, bangs, military uniform, gloves, belt, coat, blunt bangs, hat, slim body, hourglass body shape,....
Weight 1.0 CFG 7.0
SD Positive prompt (CHICHI) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of chichi, black hair, long hair, black eyes, eyelashes, blue bikini armor, pink headwear, shoulder armor, helmet, pink gloves, navel, groin, cleavage, large breasts, slim body, hourglass body shape, big thighs,....
Weight 1.0 CFG 7.0
SD Positive prompt (BULMA Young) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of bulma, dragon ball, blue hair, parted bangs, hair ribbon, braided ponytail, pink shirt, belt, scarf, pink skirt, brown gloves, blue eyes,..
SD Positive prompt (BULMA Adult) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of bulma, dragon ball, aqua hair, very short hair, earrings, jewelry, red dress, yellow scarf, short dress, sleeveless, blue eyes,,..
SD1.5 Weight 0.95-1.0 CFG 7.0
SD Positive prompt (C18) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of android c18, blue short-sleeved denim jacket with a blue collared and white shirt with long sleeves,cleavage, brown belt, blue eyes, collarbone, denim jacket, earrings, jewelry, lips, long sleeves, white shirt, blonde hair, short hair,..
SD1.5 Weight 0.95-1.0 CFG 7.0
SD Positive prompt (Good LAUNCH) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of launch, blue hair, massive cleavage, green crop top, navel, short shorts, wavy hair, long hair, hairband, bare shoulders, parted bangs, big breasts, blue eyes, slim body, hourglass body shape,big thighs, long eyelashes,..
SD Positive prompt (Bad LAUNCH) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of launch, blonde hair, massive cleavage, gloves, green crop top, navel, short shorts, wavy hair, very long hair, hairband, bare shoulders, parted bangs, big breasts, blue eyes, slim body, hourglass body shape,big thighs, long eyelashes,..
SD1.5 Weight 0.95-1.0 CFG 7.0
SD Positive prompt (PAN DBGT) : LORA, realistic, masterpiece, realistic, detailed, 8k, 1girl, beautiful photography of pan, black eyes, short hair, black hair, red shirt, crop top, orange bandana, black gloves, parted bangs, short pants, long eyelashes, slim body, hourglass body shape,big thighs, petite girl,,,..
SD1.5 Weight 0.9-0.95 CFG 7.0
For perfect results, use only suggested embedding in negative prompt and suggested upscaler. Don't use with other loras and ng_deepnegative_v1_75t embedding (use only for anime or semirealistic) in negative prompt. In this way you have realistic body and skin detail.
Recommended for a better result: ADETAILER mediapipe face mesh with impaint 0.55
Setting : Sampling DPM Karras, HiresFix on, ITF-skinDiffDetail-Lite at 2x, Denoising 0.38, Upscale 2x, Sampling steps 35+, Resolution 512x768, Upscale 2x ITF-skinDiffDetail-Lite next passage.
Not for sell.
If you appreciate the lora, please consider leaving us a review and sharing the images generated using this model. Your feedback is valuable to both us and the AI community. Thanks.
Description
See info
FAQ
Comments (4)
You covered what settings to use (much appreciated!)...all except the checkpoint. What checkpoint do you recommend?
I see the "Epic Girls" checkpoint as a "Suggested Resource". Is that the recommended checkpoint? I ask only because it looks a bit more 2D than your example images. Just wanted to be sure.
First, you need to use a realistic checkpoint. You have to put in the prompt, at the beginning, (realistic:1.3),(photorealistic) and put in negative (anime, 3d rendering). I use my checkpoints and share them on TensorArt, in this case I use Realistic2Epic enhanced. Everything else is explained in the description. If you don't have realistic results, it depends a lot on the prompt you used and whether you used embedding or other loras, along with the main lora.
@MyIA Thank you for the reply. Results also depends on the checkpoint used, and how far its weights have deviated from the checkpoint these LoRAs were trained on.
After thoroughly testing, photorealistic checkpoints don't play nice with these. They fudge unnatural hair colors, place bias on a large thick muscular body shape with huge hips & massive thighs, and they're extremely biased towards certain poses (turned to the side w/ hands on hips or head down, looking up), regardless of prompts, CFG, and LoRA strength. They were tested with epicphotogasm_x, analogmadness, aZoyaPhotoreal, wafflemix, photon, epicrealism, realisticvision, and a number of others.
Conversely, there are a few 2.5D/semi-realistic models that work great with these...
chilloutmix_NiPrunedFp32 (clip skip 1)
fcblendmixfcRealism_v21 (clip skip 1)
animerge_v24 (clip skip 2)
TheTrualityEngine (clip skip 1)
my own personal checkpoint (clip skip 1)
They produce great results! However, they still show bias towards large manly muscular bodies with over-exaggerated hips, but this can be handled with neg prompts, slider LoRAs, and/or using a 0.7-0.8 LoRA strength w/ aDetailer face_yolov8n (0.55 denoise) fixing the face using a photoreal checkpoint (such as epicphotogasm_x).
Great work on these!
@Bit_Shifter It's partly true. I have made several realistic models and they work well. The only one I left on Civitai is EpicGirls. In all SD models I use Clip a 2 not a 1. In addition, I have indicated the 3 negative embeddings I use. I don't know the lora trained on a mix of anime and real photos, if a checkpoint is not trained on everything and is not censored, it is not able to reproduce the lora respecting the prompt. With my checkpoints you have to avoid comfyui since they were not made to work with that but using A1111. You may have different results. Try EpicGirl with my prompts and putting Clip 2 and see if you have good results. Unfortunately I don't think I will share any more checkpoints on CivitAi until they change their discriminatory policy on those who are not subscribers to the site. Adetailer face_yolov8n (0.55 denoise) is wrong because change face, try face mesh all. This is the my checkpoint with i train and realize SD realistic loras : https://tensor.art/models/795064997278504434
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.