[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86343-en":3,"doc-seo-86343-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86343,7971461741311,"Ophelia","https://ap-avatar.wpscdn.com/avatar/74000253aff267980c6?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779345379180704826",8,"Research & Report","Latent-Identity Tuning in Text-to-Image Personalization Models","High-precision personalization and editing of human faces are challenging because even small changes can shift perceived identity, while existing text-to-image methods often lack fine control. This work introduces a latent identity tuning approach that directly modifies identity tokens inside a pretrained, frozen text-to-image personalization encoder to enable localized, fine-grained, and semantically coherent facial attribute edits. The tuned identity remains consistent across prompts and scenes, verified through qualitative and quantitative experiments.","arXiv :2607 . 11885v1 [ cs .CV] 13 Jul 2026  \nLatent-Identity Tuning in Text-to-Image Personalization Models  \nDaniel Garibi 1 Ronen Kamenetsky 1 Hadar Averbuch-Elor2 Daniel Cohen-Or1 Or Patashnik1  \n1Tel Aviv University 2 Cornell University  \n[https://garibida.github.io/IdentityTuning/](https://garibida.github.io/IdentityTuning/)  \nModify Nose  \nAdd Freckles  \nAdd Beard  \nFigure 1 . We present methods for directly tuning the identity tokens of a personalization encoder, enabling fine-grained control of facial attributes, for example modifying the nose, adding freckles, or a beard (top) . The edited identity can then be used across diverse prompts to generate the same tuned subject consistently in new scenes (bottom) .  \nAbstract regions. We show that meaningful directions can be iden  \nGenerating and editing a person’s face demands high tified within this space and within subspaces defined by seprecision, as even minor modifications can significantly lected tokens, enabling localized, fine-grained, and semanalter a subject’s perceived identity. Current personaliza- tically coherent edits. We validate our approach throughtion and editing methods built on general-purpose text-to- qualitative and quantitative experiments that demonstrate image models, however, often lack the precision required diverse localized facial edits while preserving cross-image for fine-grained facial edits. We present a method for identity consistency.  \nfine-grained identity tuning in text-to-image personalization  \nmodels. Unlike standard image editing, which operates on 1. Introduction  \na given image, identity tuning modifies the latent represen- Advances in text-to-image generation [33, 46, 61, 67] have tation of a specific identity, enabling the generation of di- enabled highly personalized synthesis, where models can verse images that consistently depict the same edited iden- generate new images of a person from a single or a few tity. To enable fine-grained latent identity tuning, we ex- reference images. Such personalization methods allow theplore the latent space of a pre-trained, frozen encoder for depiction of an individual in various contexts that vary in text-to-image personalization. Our approach requires no background, style, and pose [19, 40, 63, 77] . However, additional training. Instead, it leverages the existing archi- while existing methods faithfully reproduce identity, they tecture of a frozen encoder to uncover latent semantic direc- offer little control over modifying it, limiting users’ ability tions. This space consists of a set of latent tokens that play to refine, customize, or creatively reinterpret how a given distinct roles in capturing different aspects of an identity identity is depicted. For instance, a person may wish to apand often correspond to specific spatial or semantic facial pear with a beard, freckles, or a modified nose, requiring  \nthis edited appearance to remain consistent across all subsequent generations (Fig. 1) .  \nWe define identity tuning as the task of modifying the latent representation that encodes a person’s identity in a pretrained text-to-image personalization model. In contrast to image editing, which manipulates individual images, identity tuning operates on the latent identity representation itself. This ensures that the modified attributes remain consistent across all generated samples, establishing a persistent, altered identity that can be reliably deployed across varied prompts and settings [17, 52] . As this task involves human faces, it requires exceptional precision, as even minute differences can alter the perception of the identity. Achieving this level of precision requires enabling localized and continuous control over how the identity is portrayed or intended to appear.  \nExisting personalization methods typically learn an embedding or a set of latent tokens that capture a person’s identity from one or more reference images [29, 59, 77] . While these representations are primaril","cbCaij8KVoFBhAfG","https://ap.wps.com/l/cbCaij8KVoFBhAfG","pdf",27334774,4,1,19,"English","en",105,"# Abstract\n# Introduction\n## Identity tuning definition\n## Limitations of existing personalization methods\n## Latent space analysis and editing directions\n# Framework effectiveness","[{\"question\":\"What is latent-identity tuning in text-to-image personalization models?\",\"answer\":\"It modifies the latent representation that encodes a person’s identity by directly tuning identity tokens inside a pretrained personalization encoder, rather than editing individual images.\"},{\"question\":\"How does the method achieve fine-grained and localized facial edits?\",\"answer\":\"By uncovering meaningful latent directions in the identity token space and identifying tokens most relevant to specific localized facial attributes, enabling targeted continuous edits.\"},{\"question\":\"How is identity consistency maintained across different prompts and scenes?\",\"answer\":\"The approach produces a tuned latent identity representation that can be reused when generating new images from diverse prompts, yielding consistent edited identity features.\"}]",1784210662,48,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"latent-identity-tuning-in-text-to-image-personalization-models","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/latent-identity-tuning-in-text-to-image-personalization-models/86343/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-26","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is latent-identity tuning in text-to-image personalization models?","Question",{"text":75,"@type":76},"It modifies the latent representation that encodes a person’s identity by directly tuning identity tokens inside a pretrained personalization encoder, rather than editing individual images.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the method achieve fine-grained and localized facial edits?",{"text":80,"@type":76},"By uncovering meaningful latent directions in the identity token space and identifying tokens most relevant to specific localized facial attributes, enabling targeted continuous edits.",{"name":82,"@type":73,"acceptedAnswer":83},"How is identity consistency maintained across different prompts and scenes?",{"text":84,"@type":76},"The approach produces a tuned latent identity representation that can be reused when generating new images from diverse prompts, yielding consistent edited identity features.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},"General","general"]