arXiv:2508.06937v2 Announce Type: replace 
Abstract: Recent advances in text-to-image (T2I) models have enabled training-free regional image editing by leveraging the generative priors of foundation models. However, existing methods struggle to balance text adherence in edited regions, context fidelity in unedited areas, and seamless integration of edits. We introduce CannyEdit, a novel training-free framework that addresses this trilemma through two key innovations. First, Selective Canny Control applies structural guidance from a Canny ControlNet only to the unedited regions, preserving the original image's details while allowing for precise, text-driven changes in the specified editable area. Second, Dual-Prompt Guidance utilizes both a local prompt for the specific edit and a global prompt for overall scene coherence. Through this synergistic approach, these components enable controllable local editing for object addition, replacement, and removal, achieving a superior trade-off among text adherence, context fidelity, and editing seamlessness compared to current region-based methods. Beyond this, CannyEdit offers exceptional flexibility: it operates effectively with rough masks or even single-point hints in addition tasks. Furthermore, the framework can seamlessly integrate with vision-language models in a training-free manner for complex instruction-based editing that requires planning and reasoning. Our extensive evaluations demonstrate CannyEdit's strong performance against leading instruction-based editors in complex object addition scenarios.

CannyEdit هو إطار مبتكر يحدث ثورة في تحرير الصور من خلال السماح للمستخدمين بإجراء تعديلات انتقائية دون الحاجة إلى تدريب مكثف. هذه الابتكار مهم لأنه يوازن بشكل فعال بين الحاجة إلى الالتزام بالنص في المناطق المعدلة مع الحفاظ على سلامة الأقسام غير المعدلة. من خلال معالجة التحديات الشائعة في تحرير الصور، يفتح CannyEdit آفاقًا جديدة للمبدعين ويعزز تجربة التحرير بشكل عام.

CannyEdit es un marco innovador que revoluciona la edición de imágenes al permitir a los usuarios realizar ediciones selectivas sin necesidad de un entrenamiento extenso. Esta innovación es significativa porque equilibra de manera efectiva la necesidad de adherencia al texto en las áreas editadas mientras mantiene la integridad de las secciones no editadas. Al abordar los desafíos comunes en la edición de imágenes, CannyEdit abre nuevas posibilidades para los creadores y mejora la experiencia general de edición.

CannyEdit est un cadre révolutionnaire qui transforme l'édition d'images en permettant aux utilisateurs de faire des modifications sélectives sans nécessiter de formation approfondie. Cette innovation est importante car elle équilibre efficacement le besoin d'adhérence au texte dans les zones modifiées tout en préservant l'intégrité des sections non modifiées. En abordant les défis courants de l'édition d'images, CannyEdit ouvre de nouvelles possibilités pour les créateurs et améliore l'expérience globale d'édition.

CannyEdit is a groundbreaking framework that revolutionizes image editing by allowing users to make selective edits without the need for extensive training. This innovation is significant because it effectively balances the need for text adherence in edited areas while maintaining the integrity of unedited sections. By addressing common challenges in image editing, CannyEdit opens up new possibilities for creators and enhances the overall editing experience.

CannyEdit: Selective Canny Control and Dual-Prompt Guidance for Training-Free Image Editing

arXiv:2511.18684v1 Announce Type: new 
Abstract: Robust concept removal for text-to-image (T2I) and text-to-video (T2V) models is essential for their safe deployment. Existing methods, however, suffer from costly retraining, inference overhead, or vulnerability to adversarial attacks. Crucially, they rarely model the latent semantic overlap between the target erase concept and surrounding content -- causing collateral damage post-erasure -- and even fewer methods work reliably across both T2I and T2V domains. We introduce Instant Concept Erasure (ICE), a training-free, modality-agnostic, one-shot weight modification approach that achieves precise, persistent unlearning with zero overhead. ICE defines erase and preserve subspaces using anisotropic energy-weighted scaling, then explicitly regularises against their intersection using a unique, closed-form overlap projector. We pose a convex and Lipschitz-bounded Spectral Unlearning Objective, balancing erasure fidelity and intersection preservation, that admits a stable and unique analytical solution. This solution defines a dissociation operator that is translated to the model's text-conditioning layers, making the edit permanent and runtime-free. Across targeted removals of artistic styles, objects, identities, and explicit content, ICE efficiently achieves strong erasure with improved robustness to red-teaming, all while causing only minimal degradation of original generative abilities in both T2I and T2V models.

قدم الباحثون تقنية Instant Concept Erasure (ICE)، وهي نهج جديد لإزالة المفاهيم بشكل موثوق في نماذج النص إلى صورة (T2I) والنص إلى فيديو (T2V). تلغي هذه الطريقة الحاجة إلى إعادة التدريب المكلفة وتقلل من عبء الاستدلال، بينما تعالج الثغرات الأمنية ضد الهجمات العدائية. تستخدم ICE تقنية تعديل الوزن في خطوة واحدة دون تدريب، مما يضمن إلغاء تعلم دقيق ودائم دون أضرار جانبية للمحتوى المحيط.

Investigadores han presentado el Instant Concept Erasure (ICE), un enfoque novedoso para la eliminación robusta de conceptos en modelos de texto-a-imagen (T2I) y texto-a-video (T2V). Este método elimina la necesidad de reentrenamiento costoso y minimiza la sobrecarga de inferencia, al tiempo que aborda las vulnerabilidades a ataques adversariales. ICE emplea una técnica de modificación de pesos en una sola vez, sin necesidad de entrenamiento, que asegura un desaprendizaje preciso y persistente sin daños colaterales al contenido circundante.

Des chercheurs ont introduit l'Instant Concept Erasure (ICE), une nouvelle approche pour la suppression robuste de concepts dans les modèles de texte-à-image (T2I) et de texte-à-vidéo (T2V). Cette méthode élimine le besoin de réentraînement coûteux et minimise la surcharge d'inférence tout en abordant les vulnérabilités aux attaques adversariales. ICE utilise une technique de modification de poids en une seule fois, sans formation, qui garantit un désapprentissage précis et persistant sans dommages collatéraux au contenu environnant.

Researchers have introduced Instant Concept Erasure (ICE), a novel approach for robust concept removal in text-to-image (T2I) and text-to-video (T2V) models. This method eliminates the need for costly retraining and minimizes inference overhead while addressing vulnerabilities to adversarial attacks. ICE employs a training-free, one-shot weight modification technique that ensures precise and persistent unlearning without collateral damage to surrounding content.

Now You See It, Now You Don't - Instant Concept Erasure for Safe Text-to-Image and Video Generation

arXiv:2511.18131v1 Announce Type: new 
Abstract: We observe that recent advances in multimodal foundation models have propelled instruction-driven image generation and editing into a genuinely cross-modal, cooperative regime. Nevertheless, state-of-the-art editing pipelines remain costly: beyond training large diffusion/flow models, they require curating massive high-quality triplets of \{instruction, source image, edited image\} to cover diverse user intents. Moreover, the fidelity of visual replacements hinges on how precisely the instruction references the target semantics. We revisit this challenge through the lens of temporal modeling: if video can be regarded as a full temporal process, then image editing can be seen as a degenerate temporal process. This perspective allows us to transfer single-frame evolution priors from video pre-training, enabling a highly data-efficient fine-tuning regime. Empirically, our approach matches the performance of leading open-source baselines while using only about one percent of the supervision demanded by mainstream editing models.

أدت التطورات الأخيرة في النماذج متعددة الوسائط إلى منظور جديد حول تحرير الصور، حيث يتم اعتباره عملية زمنية متدهورة. تتيح هذه الطريقة نقل أولويات تطور الإطار الواحد من التدريب على الفيديو، مما يعزز كفاءة البيانات في ضبط نماذج تحرير الصور. تتطابق الطريقة مع أداء أفضل الأسس مفتوحة المصدر بينما تقلل من الحاجة إلى مجموعات بيانات ثلاثية واسعة.

Los avances recientes en modelos multimodales han llevado a una nueva perspectiva sobre la edición de imágenes, considerándola como un proceso temporal degenerado. Este enfoque permite la transferencia de prioridades de evolución de un solo fotograma a partir del preentrenamiento de video, mejorando la eficiencia de los datos en el ajuste fino de los modelos de edición de imágenes. El método iguala el rendimiento de las principales bases de código abiertas mientras reduce la necesidad de conjuntos de datos tripletes extensos.

Des avancées récentes dans les modèles multimodaux ont conduit à une nouvelle perspective sur l'édition d'images, la considérant comme un processus temporel dégénéré. Cette approche permet le transfert de priorités d'évolution d'images uniques issues de l'entraînement vidéo, améliorant ainsi l'efficacité des données dans le réglage fin des modèles d'édition d'images. La méthode égalise les performances des meilleures bases de code ouvertes tout en réduisant le besoin de jeux de données triplets étendus.

Recent advancements in multimodal foundation models have led to a new perspective on image editing, viewing it as a degenerate temporal process. This approach allows for the transfer of single-frame evolution priors from video pre-training, enhancing data efficiency in fine-tuning image editing models. The method matches the performance of leading open-source baselines while reducing the need for extensive curated datasets.

Video4Edit: Viewing Image Editing as a Degenerate Temporal Process

arXiv:2505.11579v2 Announce Type: replace-cross 
Abstract: As AI systems evolve from static tools to dynamic agents, traditional categorical governance frameworks -- based on fixed risk tiers, levels of autonomy, or human oversight models -- are increasingly insufficient on their own. Systems built on foundation models, self-supervised learning, and multi-agent architectures increasingly blur the boundaries that categories were designed to police. In this Perspective, we make the case for dimensional governance: a framework that tracks how decision authority, process autonomy, and accountability (the 3As) distribute dynamically across human-AI relationships. A critical advantage of this approach is its ability to explicitly monitor system movement toward and across key governance thresholds, enabling preemptive adjustments before risks materialize. This dimensional approach provides the necessary foundation for more adaptive categorization, enabling thresholds and classifications that can evolve with emerging capabilities. While categories remain essential for decision-making, building them upon dimensional foundations allows for context-specific adaptability and stakeholder-responsive governance that static approaches cannot achieve. We outline key dimensions, critical trust thresholds, and practical examples illustrating where rigid categorical frameworks fail -- and where a dimensional mindset could offer a more resilient and future-proof path forward for both governance and innovation at the frontier of artificial intelligence.

تطور أنظمة الذكاء الاصطناعي من أدوات ثابتة إلى وكلاء ديناميكيين يتطلب تغييرًا في أطر الحوكمة، حيث أصبحت النماذج الفئوية التقليدية غير كافية بشكل متزايد. يركز إطار الحوكمة البُعدية المقترح على التوزيع الديناميكي لسلطة القرار، واستقلالية العمليات، والمساءلة في العلاقات بين الإنسان والذكاء الاصطناعي، بهدف معالجة المخاطر بشكل استباقي قبل أن تتجلى.

La evolución de los sistemas de IA de herramientas estáticas a agentes dinámicos requiere un cambio en los marcos de gobernanza, ya que los modelos categóricos tradicionales son cada vez más inadecuados. El marco de gobernanza dimensional propuesto se centra en la distribución dinámica de la autoridad de decisión, la autonomía del proceso y la responsabilidad en las relaciones humano-IA, con el objetivo de abordar proactivamente los riesgos antes de que se materialicen.

L'évolution des systèmes d'IA, passant d'outils statiques à des agents dynamiques, nécessite un changement dans les cadres de gouvernance, car les modèles catégoriels traditionnels deviennent de plus en plus inadéquats. Le cadre de gouvernance dimensionnelle proposé se concentre sur la distribution dynamique de l'autorité décisionnelle, de l'autonomie des processus et de la responsabilité dans les relations homme-IA, visant à anticiper les risques avant qu'ils ne se matérialisent.

The evolution of AI systems from static tools to dynamic agents necessitates a shift in governance frameworks, as traditional categorical models are increasingly inadequate. The proposed dimensional governance framework focuses on the dynamic distribution of decision authority, process autonomy, and accountability in human-AI relationships, aiming to preemptively address risks before they materialize.

CannyEdit: Selective Canny Control and Dual-Prompt Guidance for Training-Free Image Editing

Was this article worth reading? Share it

IMGFX.DEV

Golan AI

PicPicAI