Adding Object into images based on text instructions is a challenging task in semantic image editing, requiring a balance between preserving the original scene and seamlessly integrating the new object in a fitting location. Despite extensive efforts, existing models often struggle with this balance, particularly with finding a natural location for adding an object in complex scenes. We introduce Add-it, a training-free approach that extends diffusion models' attention mechanisms to incorporate information from three key sources: the scene image, the text prompt, and the generated image itself. Our weighted extended-attention mechanism maintains structural consistency and fine details while ensuring natural object placement. Without task-specific fine-tuning, Add-it achieves state-of-the-art results on both real and generated image insertion benchmarks, including our newly constructed "Additing Affordance Benchmark" for evaluating object placement plausibility, outperforming supervised methods. Human evaluations show that Add-it is preferred in over 80% of cases, and it also demonstrates improvements in various automated metrics.

Michael Toker, Ido Galil, Hadas Orgad, Rinon Gal, Yoad Tewel, Gal Chechik, Yonatan Belinkov. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Text-to-image generation has evolved beyond single monolithic models to complex multi-component pipelines that combine various enhancement tools. While these pipelines significantly improve image quality, their effective design requires substantial expertise. Recent approaches automating this process through large language models (LLMs) have shown promise but suffer from two critical limitations: extensive computational requirements from generating images with hundreds of predefined pipelines, and poor generalization beyond memorized training examples. We introduce a novel reinforcement learning-based framework that addresses these inefficiencies. Our approach first trains an ensemble of reward models capable of predicting image quality scores directly from prompt-workflow combinations, eliminating the need for costly image generation during training. We then implement a two-phase training strategy: initial workflow prediction training followed by GRPO-based optimization that guides the model toward higher-performing regions of the workflow space. Additionally, we incorporate a classifier-free guidance based enhancement technique that extrapolates along the path between the initial and GRPO-tuned models, further improving output quality. We validate our approach through a set of comparisons, showing that it can successfully create new flows with greater diversity and lead to superior image quality compared to existing baselines.

A sketch is one of the most intuitive and versatile tools humans use to convey their ideas visually. An animated sketch opens another dimension to the expression of ideas and is widely used by designers for a variety of purposes. Animating sketches is a laborious process, requiring extensive experience and professional design skills. In this work, we present a method that automatically adds motion to a single-subject sketch (hence, “breathing life into it”), merely by providing a text prompt indicating the desired motion. The output is a short animation provided in vector representation, which can be easily edited. Our method does not require extensive training, but instead leverages the motion prior of a large pretrained text-to- video diffusion model using a score-distillation loss to guide the placement of strokes. To promote natural and smooth motion and to better preserve the sketch's appearance, we model the learned motion through two components. The first governs small local deformations and the second controls global affine transformations. Surprisingly, wefind that even models that struggle to generate sketch videos on their own can still serve as a useful backbone for animating abstract representations.

Text-to-image models offer a new level of creative flexibility by allowing users to guide the image generation process through natural language. However, using these models to consistently portraythe samesubject across diverse prompts remains challenging. Existing approaches fine-tune the model to teach it new words that describe specific user-provided subjects or add image conditioning to the model. These methods require lengthy persubject optimization or large-scale pre-training. Moreover, they struggle to align generated images with text prompts and face difficulties in portraying multiple subjects. Here, we presentConsiStory, atraining-freeapproach that enables consistent subject generation by sharing the internal activations of the pretrained model. We introduce a subject-driven shared attention block and correspondence-based feature injection to promote subject consistency between images. Additionally, we develop strategies to encourage layout diversity while maintaining subject consistency. We compareConsiStoryto a range of baselines, and demonstrate state-of-the-art performance on subject consistency and text alignment, without requiring a single optimization step. Finally,ConsiStorycan naturally extend to multi-subject scenarios, and even enable training-freepersonalizationfor common objects.

Text-to-image (T2I) personalization allows users to guide the creative image generation process by combining their own visual concepts in natural language prompts. Recently, encoder-based techniques have emerged as a new effective approach for T2I personalization, reducing the need for multiple images and long training times. However, most existing encoders are limited to a single-class domain, which hinders their ability to handle diverse concepts. In this work, we propose a domain-agnostic method that does not require any specialized dataset or prior information about the personalized concepts. We introduce a novel contrastive-based regularization technique to maintain high fidelity to the target concept characteristics while keeping the predicted embeddings close to editable regions of the latent space, by pushing the predicted tokens toward their nearest existing CLIP tokens. Our experimental results demonstrate the effectiveness of our approach and show how the learned tokens are more semantic than tokens predicted by unregularized models. This leads to a better representation that achieves state-of-the-art performance while being more flexible than previous methods.

Text-to-image personalization aims to teach a pre-trained diffusion model to reason about novel, user provided concepts, embedding them into new scenes guided by natural language prompts. However, current personalization approaches struggle with lengthy training times, high storage requirements or loss of identity. To overcome these limitations, we propose an encoder-based domain-tuning approach. Our key insight is that by underfitting on a large set of concepts from a given domain, we can improve generalization and create a model that is more amenable to quickly adding novel concepts from the same domain. Specifically, we employ two components: First, an encoder that takes as an input a single image of a target concept from a given domain, e.g. a specific face, and learns to map it into a word-embedding representing the concept. Second, a set of regularized weight-offsets for the text-to-image model that learn how to effectively injest additional concepts. Together, these components are used to guide the learning of unseen concepts, allowing us to personalize a model using only a single image and as few as 5 training steps --- accelerating personalization from dozens of minutes to seconds , while preserving quality. Code and trained encoders will be available at our project page.

The inversion of real images into StyleGAN's latent space is a well-studied problem. Nevertheless, applying existing approaches to real-world scenarios remains an open challenge, due to an inherent trade-off between reconstruction and editability: latent space regions which can accurately represent real images typically suffer from degraded semantic control. Recent work proposes to mitigate this trade-off by fine-tuning the generator to add the target image to well-behaved, editable regions of the latent space. While promising, this fine-tuning scheme is impractical for prevalent use as it requires a lengthy training phase for each new image. In this work, we introduce this approach into the realm of encoder-based inversion. We propose HyperStyle, a hypernetwork that learns to modulate StyleGAN's weights to faithfully express a given image in editable regions of the latent space. A naive modulation approach would require training a hypernetwork with over three billion parameters. Through careful network design, we reduce this to be in line with existing encoders. HyperStyle yields reconstructions comparable to those of optimization techniques with the near real-time inference capabilities of encoders. Lastly, we demonstrate HyperStyle's effectiveness on several applications beyond the inversion task, including the editing of out-of-domain images which were never seen during training. Code is available on our project page: https://yuval-alaluf.github.io/hyperstyle/.

We propose a novel method for solving regression tasks using few-shot or weak supervision. At the core of our method is the fundamental observation that GANs are incredibly successful at encoding semantic information within their latent space, even in a completely unsupervised setting. For modern generative frameworks, this semantic encoding manifests as smooth, linear directions which affect image attributes in a disentangled manner. These directions have been widely used in GAN-based image editing. In this work, we leverage them for few-shot regression. Specifically, we make the simple observation that distances traversed along such directions are good features for downstream tasks - reliably gauging the magnitude of a property in an image. In the absence of explicit supervision, we use these distances to solve tasks such as sorting a collection of images, and ordinal regression. With a few labels - as little as two - we calibrate these distances to real-world values and convert a pre-trained GAN into a state-of-the-art few-shot regression model. This enables solving regression tasks on datasets and attributes which are difficult to produce quality supervision for. Extensive experimental evaluations demonstrate that our method can be applied across a wide range of domains, leverage multiple latent direction discovery frame-works, and achieve state-of-the-art results in few-shot and low-supervision settings, even when compared to methods designed to tackle a single task. Code is available on our project website.

The ability of Generative Adversarial Networks to encode rich semantics within their latent space has been widely adopted for facial image editing. However, replicating their success with videos has proven challenging. Applying StyleGAN editing to real videos introduces two main challenges: (i) StyleGAN operates over aligned crops. When editing videos, these crops need to be pasted back into the frame, resulting in a spatial inconsistency. (ii) Videos introduce a fundamental barrier to overcome — temporal coherency. To address the first challenge, we propose a novel stitching-tuning procedure. The generator is carefully tuned to overcome the spatial artifacts at crop borders, resulting in smooth transitions even when difficult backgrounds are involved. Turning to temporal coherence, we propose that this challenge is largely artificial. The source video is already temporally coherent, and deviations arise in part due to careless treatment of individual components in the editing pipeline. We leverage the natural alignment of StyleGAN and the tendency of neural networks to learn low-frequency functions, and demonstrate that they provide a strongly consistent prior. These components are combined in an end-to-end framework for semantic editing of facial videos. We compare our pipeline to the current state-of-the-art and demonstrate significant improvements. Our method produces meaningful manipulations and maintains greater spatial and temporal consistency, even on challenging talking head videos which current methods struggle with. Our code and videos are available at https://stitch-time.github.io/.

Abstract Generative Adversarial Networks (GANs) have established themselves as a prevalent approach to image synthesis. Of these, StyleGAN offers a fascinating case study, owing to its remarkable visual quality and an ability to support a large array of downstream tasks. This state‐of‐the‐art report covers the StyleGAN architecture, and the ways it has been employed since its conception, while also analyzing its severe limitations. It aims to be of use for both newcomers, who wish to get a grasp of the field, and for more experienced readers that might benefit from seeing current research trends and existing tools laid out. Among StyleGAN's most interesting aspects is its learned latent space. Despite being learned with no supervision, it is surprisingly well‐behaved and remarkably disentangled. Combined with StyleGAN's visual quality, these properties gave rise to unparalleled editing capabilities. However, the control offered by StyleGAN is inherently limited to the generator's learned distribution, and can only be applied to images generated by StyleGAN itself. Seeking to bring StyleGAN's latent control to real‐world scenarios, the study of GAN inversion and latent space embedding has quickly gained in popularity. Meanwhile, this same study has helped shed light on the inner workings and limitations of StyleGAN. We map out StyleGAN's impressive story through these investigations, and discuss the details that have made StyleGAN the go‐to generator. We further elaborate on the visual priors StyleGAN constructs, and discuss their use in downstream discriminative tasks. Looking forward, we point out StyleGAN's limitations and speculate on current trends and promising directions for future research, such as task and target specific fine‐tuning.

Can a generative model be trained to produce images from a specific domain, guided only by a text prompt, without seeing any image? In other words: can an image generator be trained "blindly"? Leveraging the semantic power of large scale Contrastive-Language-Image-Pre-training (CLIP) models, we present a text-driven method that allows shifting a generative model to new domains, without having to collect even a single image. We show that through natural language prompts and a few minutes of training, our method can adapt a generator across a multitude of domains characterized by diverse styles and shapes. Notably, many of these modifications would be difficult or infeasible to reach with existing methods. We conduct an extensive set of experiments across a wide range of domains. These demonstrate the effectiveness of our approach, and show that our models preserve the latent-space structure that makes generative models appealing for downstream tasks. Code and videos available at: stylegan-nada.github.io/

In recent years, considerable progress has been made in the visual quality of Generative Adversarial Networks (GANs). Even so, these networks still suffer from degradation in quality for high-frequency content, stemming from a spectrally biased architecture, and similarly unfavorable loss functions. To address this issue, we present a novel general-purpose Style and WAvelet based GAN (SWAGAN) that implements progressive generation in the frequency domain. SWAGAN incorporates wavelets throughout its generator and discriminator architectures, enforcing a frequency-aware latent representation at every step of the way. This approach, designed to directly tackle the spectral bias of neural networks, yields an improvement in the ability to generate medium and high frequency content, including structures which other networks fail to learn. We demonstrate the advantage of our method by integrating it into the SyleGAN2 framework, and verifying that content generation in the wavelet domain leads to more realistic high-frequency content, even when trained for fewer iterations. Furthermore, we verify that our model's latent space retains the qualities that allow StyleGAN to serve as a basis for a multitude of editing tasks, and show that our frequency-aware approach also induces improved high-frequency performance in downstream tasks.

Graph Convolutional Networks (GCN) have been recognized as successful for processing pseudo-spatial graph representations of the underlying structure of documents. We present Cardinal Graph Convolutional Networks (CGCN), an efficient and flexible extension of GCNs with cardinal-direction awareness of spatial node arrangement. The formulation of CGCNs retains the traditional GCN permutation invariance, ensuring directional neighbors are involved in learning abstract representations, even in the absence of a proper ordering of the nodes. We show that CGCNs achieve state of the art results on an invoice information extraction task, jointly learning a word-level tagging as well as document meta-level classification and regression. We also present a new multiscale Inception-like CGCN block-layer, as well as Conv-Pool-DeConv-DePool UNet-like architecture, which increase the receptive field. We demonstrate the utility of CGCNs on private and public datasets, with respect to several baseline models: sequential LSTM, transformer classifier, non-cardinal GCNs, and an image-convolutional approach.