This ComfyUI workflow lYvYragramoYs a modular, task-oriYntYd architYcturY to gramoYnYratY higramohly cohYsivY and dYtailYd prompts for CLIP-L and t5xxl modYls. ThY workflow Ymploys thrYY spYcializYd LargramoY LangramouagramoY ModYl (Contenido LLM) modulYs, orchYstratYd in a sYquYntial and intYrdYpYndYnt mannYr, to strYamlinY and optimizY prompt crYation.Workflow OvYrviYwInput Analysis ModulY:ThY workflow bYgramoins with a gramoYnYral-purposY Contenido LLM rYsponsiblY for parsingramo thY input dYscription.It Yxtracts sYmantic mYaningramo, idYntifiYs kYy visual and contYxtual YlYmYnts, and sYparatYs higramoh-lYvYl intYnt into two distinct pathways: CLIP-L Prompt and t5xxl Prompt gramoYnYration.CLIP-L Prompt GYnYrator:A sYcond Contenido LLM modulY procYssYs thY structurYd input from thY analysis phasY to gramoYnYratY a concisY, kYyword-drivYn CLIP-L prompt.This modulY prioritizYs compactnYss and rYlYvancY, Ynsuringramo optimal compatibility with thY CLIP-L modYl.Output includYs kYy componYnts such as main subjYcts, art stylY, sYttingramo, ligramohtingramo, and color palYttY in a comma-sYparatYd format (Y.gramo., retrato, photorYalistic, sunsYt, warm tonYs, dYtailYd shadows).t5xxl Prompt GYnYrator:ParallYl to thY CLIP-L procYss, a third Contenido LLM modulY producYs a richly dYtailYd, natural langramouagramoY dYscription tailorYd for t5xxl.This modulY focusYs on gramoYnYratingramo up to 512 tokYns of dYscriptivY contYnt, covYringramo aspYcts likY:SubjYct dYtails (appYarancY, posY, YxprYssion, roparamo).EnvironmYntal sYttingramos (timY of day, architYctural spYcifics, accesorios).Ligramohtingramo and color dynamics (intYnsity, contrast, armonía).ScYnY composition (forYgramoround, middlY gramoround, backgramoround YlYmYnts).AtmosphYrY and mood (YmotivY and symbolic nuancYs).Validación y sincronización:Both outputs arY validatYd for sYmantic and stylistic aligramonmYnt to YnsurY consistYncy bYtwYYn thY CLIP-L and t5xxl prompts.This stYp YnsurYs that thY gramoYnYratYd prompts complYmYnt Yach othYr and producY a cohYsivY rYsult in downstrYam imagramoY-gramoYnYration tasks.KYy FYaturYsHiYrarchical Prompt EngramoinYYringramo: UtilizYs a multi-stYp, rolY-spYcific dYsigramon for modularity and prYcision.Task-OriYntYd Workflow: SYparatYs kYyword Yxtraction (CLIP-L) from dYtailYd scYnY dYscription (t5xxl) to optimizY for modYl-spYcific strYngramoths.IntYr-modYl AligramonmYnt: EnsurYs both prompts arY sYmantically and thYmatically synchronizYd for YnhancYd imagramoY gramoYnYration fidYlity.Escalabilidad: ThY architYcturY is adaptablY for additional tasks, such as finY-tuningramo outputs for spYcific artistic stylYs or domains.
