Latent Diffusion Models
PulseAugur coverage of Latent Diffusion Models — every cluster mentioning Latent Diffusion Models across labs, papers, and developer communities, ranked by signal.
- 2026-07-14 research_milestone Researchers introduced a new method for paragraph-level handwriting imitation using modified latent diffusion models, setting a new benchmark in style preservation. source
3 day(s) with sentiment data
-
New JiT-DDT architecture trains text-to-image models 3.6x faster
Researchers have developed JiT-DDT, a novel architecture that significantly accelerates the training of text-to-image diffusion models. By unifying the compression and generation modules into a single model, JiT-DDT ach…
-
New framework enhances Latent Diffusion Models for medical imaging analysis
Researchers have developed a new fine-tuning framework to improve the multi-modal alignment of pre-trained Latent Diffusion Models for medical imaging tasks. This approach addresses the limited data availability in medi…
-
PhysWave uses physics priors for controllable spatial audio generation
Researchers have developed PhysWave, a novel physics-guided latent diffusion model designed for controllable spatial audio generation. This model unifies natural language and trajectory control through a shared waypoint…
-
Research questions effectiveness of guidance methods in latent diffusion models
A new research paper revisits Classifier-Free Guidance (CFG) methods for latent diffusion models, evaluating eight training-free techniques on two open-weight rectified-flow transformers. The study found that no single …
-
AI image models like Stable Diffusion now run in-browser via TypeScript
New advancements in ONNX Runtime Web and WebGPU are enabling complex AI models like Stable Diffusion and Flux to run directly within web browsers using TypeScript. This shift moves AI model execution from dedicated back…
-
MiniMax AI open-sources VTP framework for generative models
MiniMax AI has open-sourced its Visual Tokenizer Pre-training (VTP) framework, which was developed to validate a scaling law for visual tokenizers. This framework is designed for next-generation generative models and ch…
-
New DeepFreqMark framework embeds watermarks in AI images
Researchers have developed DeepFreqMark, a novel framework for embedding watermarks into AI-generated images from Latent Diffusion Models (LDMs). Unlike previous methods that used fixed patterns, DeepFreqMark employs a …
-
New method enables controllable clothing generation for virtual try-on
A new method for virtual try-on applications has been developed, utilizing latent diffusion models to generate controllable clothing images. This approach augments image data with precise labels for garment lengths and …
-
Flash-VAED framework accelerates video generation by 6x
Researchers have developed Flash-VAED, a framework designed to accelerate the VAE decoders used in latent diffusion models for video generation. This approach employs channel pruning and dominant operator optimization t…
-
New AI watermarking techniques emerge amid security concerns · 3 sources tracked
Researchers have developed new methods for watermarking AI-generated images to ensure authenticity and prevent forgery. One approach, IRIS, binds watermarks to the image's visual semantics, making them resistant to tran…
-
New DiffAttack method uses diffusion models to fool face recognition systems
Researchers have developed a new method called DiffAttack that uses latent diffusion models to create adversarial examples for face recognition systems. This approach optimizes within the latent space of diffusion model…
-
Diffusion models advance medical image inpainting, survey finds
A recent survey paper published on arXiv details the advancements and challenges in using diffusion models for medical image inpainting. The paper systematically reviews 60 studies, highlighting the growing research int…
-
New research explores advanced diffusion models for generation, robustness, and speed
Researchers are developing advanced diffusion models for various applications, including image generation, time-series synthesis, and natural language processing. New methods like Simplax aim to improve categorical gene…
-
New methods enhance multimodal control in Diffusion Transformers for image generation
Researchers have developed new methods to enhance control over image generation using Diffusion Transformers (DiTs). One approach, 'Appearance Pointers,' uses compact tokens to guide DiTs for precise regional control ov…
-
New research enhances diffusion language models for efficiency and semantics · 3 sources tracked
Researchers have developed new methods to improve diffusion language models, addressing limitations in their efficiency and semantic understanding. One approach, JUMP, enhances membership inference attacks by enabling s…
-
New diffusion model achieves paragraph-level handwriting imitation
Researchers have developed a novel method for imitating handwriting at the paragraph level, overcoming the limitations of existing models that primarily generate individual words or lines. This new approach utilizes a m…
-
GMODiff framework refines HDR reconstruction using diffusion models
Researchers have developed GMODiff, a novel one-step diffusion framework for High Dynamic Range (HDR) reconstruction. This method reframes HDR reconstruction as a gain map refinement problem, leveraging pre-trained Late…
-
New theory explains forgery attacks on AI watermarks
Researchers have developed a new theoretical framework to understand the vulnerabilities of semantic watermarks in latent diffusion models (LDMs). Their analysis reveals that structural mismatches between different mode…
-
PixelU Transformer offers efficient end-to-end pixel diffusion
Researchers have introduced PixelU, a novel U-shaped Diffusion Transformer designed for efficient end-to-end pixel diffusion. This model challenges the necessity of complex decoders in pixel-space diffusion by focusing …
-
New SSVAE method boosts video diffusion model training speed by 3x
Researchers have developed a new method called Spectral-Structured VAE (SSVAE) to improve the performance of latent diffusion models used in video generation. By analyzing the latent spaces of video Variational Autoenco…