Meng-Chieh Ling

#### First Release on 2024-10-03 ####Workflow: https://tensor.art/workflows/780398465644743890Introduction:This is a workflow for creating realistic European beauty. Since I am east Asia, I am biased to the east Asia beauty standard. If you think some features will make the image closer to a European beauty, let me know.If you are interested in knowing what prompts can be used to create the images you want, you can follow this article:Generating realistic girl showing classic East Asia aesthetic with PONY, including NSFW.The prompts mentioned in the article are tested one by one. As being a scientist, it is almost an instinct to identify what is useful and what is not.The checkpoint/model is the Realistic Pony Diffusion - v3, developed by NukeA.I.This workflow does not contain any LoRA and can be served as the playground for practicing prompt engineering, which I believe is the most important part of AIGC today. Model will improving overtime, however, the skill of prompt engineering will remain the same, at least much slower than the progress of model development.Hyper Parameters:Steps Basic: number of denoise steps for the basic imageSteps Hires: number of denoise steps for the upscalingsampler: dpmpp_sde. (Suggested by NukeAI)scheduler: karrasDenoise Hires Fix: denoise factor for upscaleRandom Seed: seed for the noise generation. It remains fixed after image generation.CFG: classifier-free guidance.Aspect RatioUpscale FactorPrompts:Due to programming habits, the prompts are separated into 7 blocks. Of course you can put whatever prompts in which ever blocks you like.In the end of each prompt block, you need a comma "," so the prompts can be concatenated properly.Default. This part cannot be changed and the content is:1girl, a beautiful russian girl, score_9, score_8_up, score_7_up, source_photo, lustry skin texture, pale skin tone, photography, realistic, hyperrealistic, ultrarealistic, raw photo, absurd resolution, blue eyes, profound facial features, caucasian descendant, very large breasts,Face and Makeup: the place for face description and makeup prompt.Hair Style: the place for hair style description.Pose: the place for pose control.Outfit: the place for the outfit description.Theme: the topic of this image. Of anything not related to the girl can be put here.NSFW: the place for the NSFT prompt. What prompt can be used? Go to my article to find out.Negative prompts:score_6, score_5, score_4, source_pony, source_furry, source_cartoon, source_anime, muscular female, ugly face, ugly, verybadimagenegative_v1.3, bad-picture-chill-75v, bad_pictures, bad-hands-5, EasynegativeV2, blurry face

#### Updated on 2024-09-01 ####Workflow: https://tensor.art/workflows/780399951703419618Introduction:This is a workflow for creating realistic girl with a classical East Asian beauty standard.If you are interested in knowing what prompts can be used to create the images you want, you can follow this article:Generating realistic girl showing classic East Asia aesthetic with PONY, including NSFW.The prompts mentioned in the article are tested one by one. As being a scientist, it is almost an instinct to identify what is useful and what is not.The checkpoint/model is the Realistic Pony Diffusion - v3, developed by NukeA.I.This workflow does not contain any LoRA and can be served as the playground for practicing prompt engineering, which I believe is the most important part of AIGC today. Model will improving overtime, however, the skill of prompt engineering will remain the same, at least much slower than the progress of model development.This work flow includes a upscale step and a face ADetailer step, therefore it is quite expensive in exchange for the image quality. The upscale has three part: 4x-Ultrasharp -> Scale to 1000 * 1500 -> image generation.Hyper Parameters:Steps Basic: number of denoise steps for the basic imageSteps Hires: number of denoise steps for the upscalingsampler: dpmpp_sde. (Suggested by NukeAI)scheduler: karrasDenoise Hires Fix: denoise factor for upscaleRandom Seed: seed for the noise generation. It remains fixed after image generation.CFG: classifier-free guidance.Aspect RatioUpscale FactorPrompts:Due to programming habits, the prompts are separated into 7 blocks. Of course you can put whatever prompts in which ever blocks you like.Default. This part cannot be changed and the content is:score_9, score_8_up, score_7_up, source_photo, source_anime, 1girl, an anime-like beautiful japanese girl, ulzzang, lustry skin texture, pale skin tone, photography, realistic, hyperrealistic, ultrarealistic, absurd resolution,Face and Makeup: the place for face description and makeup prompt.Hair Style: the place for hair style description.Pose: the place for pose control.Outfit: the place for the outfit description.Theme: the topic of this image. Of anything not related to the girl can be put here.NSFW: the place for the NSFT prompt. What prompt can be used? Go to my article to find out.Negative prompts:score_6, score_5, score_4, score_3, score_2, score_1, source_pony, source_furry, source_cartoon, source_anime, western, muscular female, ugly face, ugly, verybadimagenegative_v1.3, bad-picture-chill-75v, bad_pictures, bad-hands-5, EasynegativeV2, blurry face##########In this update, the Adetailer is removed because the V3 has better performance than V2. As a result, some computation power can be saved.

This is a workflow for Amagi Live-action V2**** 2024.08.07 ****Metadata:Checkpoint: Realistic Pony Diffusion (https://tensor.art/models/739823076377766442)LoRA:Amagi (Azur Lane) - n_a0 (https://tensor.art/models/722122871905718184/Amagi-(Azur-Lane)-n_a0-1.0)PromptDefault/Background: score_9, score_8_up, score_7_up, source_photo, source_anime, photography, realistic, hyperrealistic, ultrarealistic, amagi \(azur lane\), 1girl, fox girl, fox ears, long hair, brown hair, blunt bangs, purple eyes, fox tail, multiple tails, thick eyebrows, hair ornament, absurdresPositive Prompt - Outfit: You can put the outfit information here. Just for a better prompt design.Positive Prompt - General: You can put other prompt that might be interesting here.Negative prompt (by default): score_6_up, score_5_up, score_4_up, source_pony, source_furry, source_cartoon, western, muscular female, ugly face, man, masculine. pony_negativeV2, zPDXL-neg, EasynegativeV2, bad-hands-5, verybadimagenegative_v1.3, ng_deepnegative_v1_75t, BadDreamFastNegativeV2, UnrealisticDream, bad-picture-chill-75vOutput Format: 1000 * 1500 (the portrait format)Steps Basic: number of diffusion steps for the basic imageSteps Hires: number of diffusion steps for the Hires Fixsampler: dpmpp_2m_sde_gpuscheduler: karrasDenoise Hires Fix: denoise factor for the enlarged image10 Random SeedCFGOutfits (according to the image provided by the author of the LoRA):kimono: choker, sakuramon, purple kimono, armor, red coat, open coat, wide sleeves, long sleeves, bridal gauntletsswimsuit: ribbon choker, swimsuit, white bikini, ribbon-trimmed bikini, black sash, two-tone bow, pelvic curtain, see-through, see-through sleeves, bare shoulders, gloves13. Face Prompt: prompt for the face detailerNegative prompt for the face detailer is"""score_6_up, score_5_up, score_4_up, source_pony, source_furry, source_cartoon, western, muscular female, ugly face, man, masculine. pony_negativeV2, zPDXL-neg, EasynegativeV2, bad-hands-5, verybadimagenegative_v1.3, ng_deepnegative_v1_75t, BadDreamFastNegativeV2, UnrealisticDream, bad-picture-chill-75v"""Amagi Strength: the LoRA weight

This workflow uses the SD3 Medium with the LoRA EK-Art Booster.https://tensor.art/models/751446481946364197/EK-Art-Booster-SD3-2.0The human limb is still a problem.Metadata:Checkpoint SD3 MediumPromptt5xxl positive prompt: As far as I understand it, t5xxl helps the model better understand relationship between objects and reduces problems such as color bleed.clip_l and clip_g positive prompt: clip_l and clip_g are the CLIPS used in SDXL and acting as the `translator` between text and image. So I think this is the main place where you can place your prompt.Negative prompt (by default): 2d, 3d, unreal, animate, cartoon, anime, cg, manga, bad anatomy, bad hands, three hands, three legs, bad arms, missing legs, missing arms, poorly drawn face, bad face, fused face, cloned face, worst face, three crus, extra crus, fused crus, worst feet, three feet, fused feet, fused thigh, three thigh, fused thigh, extra thigh, worst thigh, missing fingers, extra fingers, ugly fingers, long fingers, extra eyes, 2girl, amputation, disconnected limbs, fused legs,Output Format: 1000 * 1500 (the portrait format) with Hires Fixsampler: dpmpp_2mscheduler: sgm_uniformmodel sampling shift: 3.0.Steps Basic: number of diffusion steps for the basic imageSteps Hires: number of diffusion steps for the Hires FixArt Booster LoRA weight: LoRA weight for Art Booster

This is a workflow for Kronshtadt Live-action**** 2024.06.22 ****Metadata:Checkpoint: CampurSari XL - Realistic (https://tensor.art/models/707666883865957727/CampurSari-XL-Realistic)LoRA:KronstadtXL AzurLane (https://tensor.art/models/717472998577333902/KronstadtXL-AzurLane-KronXl)PromptDefault/Background: 1girl, blonde hair, long hair, blue eyes, ultra realistic, realistic photoPositive Prompt - Outfit: You can put the outfit information here. Just for a better prompt design.Positive Prompt - General: You can put other prompt that might be interesting here.Negative prompt (by default): EasyNegative, badhandsv5-neg, verybadimagenegative_v1.3, ng_deepnegative_v1_75Output Format: 1000 * 1500 (the portrait format)Steps Basic: number of diffusion steps for the basic imageSteps Hires: number of diffusion steps for the Hires Fixsampler: dpmpp_2m_sde_gpuscheduler: karrasDenoise Hires Fix: denoise factor for the enlarged image10 Random SeedCFGOutfits (according to the image provided by the author of the LoRA):KronstadtAL: KronstadtAL, low twintails, blue eyes, dress, fur-trimmed coat, thighband pantyhose, pantyhoseKronPolice: KronPolice, shirt, bra, pencil skirt, walkie talkie, headset, police bade, handcuff, thighband pantyhose, black gloves,KronPaw: KronPaw, nipple_pasties, pasties, pantie, thighhighs ,fake cat ear, fake cat tail, paw glovesKronQipao: KronQipao, see-through, golden trim, china dress, coat, pelvic curtain, lace pantie, thighhighs, garter strapOptional: large breasts, cleavage*** The FaceDetailer is there, as a part of the pipeline but it seems not work as it should.**** 2024.06.23 ****FaceDetailer is working now. Thanks to the assistance from TA.Face Prompt: prompt for the face detailerNegataive prompt for the face detailer is"""EasyNegative, badhandsv5-neg, verybadimagenegative_v1.3, ng_deepnegative_v1_7, 3d, render, cgi, doll, cartoon, painting, illustration, fake, ugly face, 3d modeling"""*** Some of the prompt in the negative prompt seems to be blocked by the system, please check the workflow for the detail if you are interested.

Workflow MetadataCheckpoint: ToxicEchoXL - v1LoRa:Kashino - Azur Lane(https://tensor.art/models/731993419716595105)Charcoal Art XL (https://tensor.art/models/715729374999207864)Positive Prompt - Outfit:I would suggest putting the outfit prompt here, according to the creater of LoRA, Kashino has four outfits, and I will list them here so you do not have to jumping back and force between the webpages.base outfit: white military uniform, double-breasted, aiguillette, hair ribbon, buttons, cropped white jacket, purple necktie, short pleated purple skirt, black thighhighs, zettai ryouiki,optional: button gap, open clothesHot Springs Relaxation (bikini): white bikini, bikini ribbons, criss-cross halter,hair flower, multi-strapped bikini bottomoptional: crossed bandaids, bandaid on head, detached sleeves, see-through sleeves (banndaid doesn't work very well)Maid for Mayhem: white apron, black detached sleeves, frills, neck bell, clothing cutout, white bow, pleated skirt, two-toned skirt, black skirt, red lined skirt, underboob cutout, sideboob cutout,optional: hip vent (not tagged in the dataset so doesn't work super well)Robust Floral Strings (Kimono): purple kimono, japanese clothes, cleavage, off-shoulder, pelvic curtain, hair flower, hair ornamentoptional: instrument, shamisenPositive Prompt - General:The additional prompt you want to havePositive Prompt (already included as the background):KAL, charcoal \(medium\), score_9_up, score_8_up, score_7_up, score_6_up, brown hair, long hair, purple eyes, cow ears, cow horns, cow girlThe Negative prompt (already included as the background):"""bad anatomy, bad hands, three hands, three legs, bad arms, missing legs, missing arms, poorly drawn face, bad face, fused face, cloned face, worst face, three crus, extra crus, fused crus, worst feet, three feet, fused feet, fused thigh, three thigh, fused thigh, extra thigh, worst thigh, missing fingers, extra fingers, ugly fingers, long fingers, extra eyes, 2girl, amputation, disconnected limbs, fused legs,fused feet, fused thigh, three thigh, fused thigh, extra thigh, worst thigh, missing fingers, extra fingers, ugly fingers, long fingers, extra eyes, 2girl, amputation, disconnected limbs, fused legs, extra digits, fewer digits, cropped, worst quality, low quality, normal quality jpeg artifacts, signature, watermark, username, blurry, artistname"""cfgCharcoal Art Strength: the LoRA weightKashino Strength: the LoRA weight10 Other paramsters:steps: 25 (for both the base KSampler and Hires Fix KSampler)sampler: dpm++_2m_sde (for both the base KSampler and Hires Fix KSampler)scheduler: karras (for both the base KSampler and Hires Fix KSampler)Hires Fix KSampler denoise: 0.5The size of the final output is 1000 * 1500 (portrait format)

This is my first SD3 attempt with ComfyUIMetadata:Checkpoint SD3 MediumPromptDefault: masterpiece, best quality, ultra realistic, raw photo, award-winning photography, high resolution, professional portrait photographyPositive Prompt - Clip: the prompt that will be processed by Clip_L and Clip_GPositive Prompt - t5xxl: the prompt that will be processed by t5xxlNegative prompt (by default): 2d, 3d, unreal, animate, cartoon, anime, cg, manga, bad anatomy, bad hands, three hands, three legs, bad arms, missing legs, missing arms, poorly drawn face, bad face, fused face, cloned face, worst face, three crus, extra crus, fused crus, worst feet, three feet, fused feet, fused thigh, three thigh, fused thigh, extra thigh, worst thigh, missing fingers, extra fingers, ugly fingers, long fingers, extra eyes, 2girl, amputation, disconnected limbs, fused legs,Output Format: 768 * 1152 (the portrait format)Steps: 25sampler: dpmpp_2mscheduler: sgm_uniformNo Hires Fix, which will be included later**** 2024.06.22 ****An Hires Fix is addedSteps Basic: number of diffusion steps for the basic imageSteps Hires: number of diffusion steps for the Hires FixOutput Format: 1000 * 1500 (the portrait format)Denoise Hires Fix: denoise factor for the enlarged imagePositive Prompt - Clip (default/background): masterpiece, best quality, ultra realistic, raw photo, award-winning photography, high resolution, professional portrait photography**** 2024.08.11 ****The Previous Positive Prompt - Clip is now renamed as SDXL Prompt. Because clip_l and clip_g are the CLIPS used in SDXL and acting as the `translator` between text and image. So I think this is the main place where you can place your prompt. The prompt are separated into three parts so have a better control over the content.Random seed control will be added in the next version.

Workflow MetadataCheckpoint: ToxicEtheRealXL - v1LoRa: NIKKE: Goddess of Victory Profile | Concept LoRA XL (https://tensor.art/models/683091965761332743)stellar blade tachy (https://tensor.art/models/736868864727675793)Promptfollowing prompts are already included:Positive Prompt - Tachy: The following prompts are already included:"""score_9, score_8_up, score_7_up, stellar_blade_tachy, masterpiece, best quality, uhd, perfect details, intricate details, high resolution, official art"""Positive Prompt - Nikke Profile:The following prompts are already included"""IncrsNikkeProfile, zoom layer"""The Negative prompt:"""bad anatomy, bad hands, three hands, three legs, bad arms, missing legs, missing arms, poorly drawn face, bad face, fused face, cloned face, worst face, three crus, extra crus, fused crus, worst feet, three feet, fused feet, fused thigh, three thigh, fused thigh, extra thigh, worst thigh, missing fingers, extra fingers, ugly fingers, long fingers, extra eyes, 2girl, amputation, disconnected limbs, fused legs,fused feet, fused thigh, three thigh, fused thigh, extra thigh, worst thigh, missing fingers, extra fingers, ugly fingers, long fingers, extra eyes, 2girl, amputation, disconnected limbs, fused legs, extra digits, fewer digits, cropped, worst quality, low quality, normal quality jpeg artifacts, signature, watermark, username, blurry, artistname"""