The technical heart of this transformation lies in how diffusion models work. Unlike earlier generative systems that produced images in a single step, diffusion models learn to reverse a gradual noising process, starting from pure randomness and iteratively denoising their way toward coherent structure. Denoising Diffusion Probabilistic Models, introduced in 2020, demonstrated that this stochastic procedure could outperform generative adversarial networks on image synthesis, and a wave of refinements followed. Most modern systems operate not on raw pixels but in a compressed latent space produced by a variational autoencoder, a strategy popularized by Latent Diffusion Models and their descendants Stable Diffusion and SDXL. This compression dramatically reduces computational cost while preserving fidelity, making high-resolution generation practical on consumer hardware.

Architecturally, the field has since moved beyond the convolutional U-Net backbone that defined the first generation of diffusion systems. The Diffusion Transformer, or DiT, replaces the U-Net with a scalable transformer and underpins flagship text-to-image systems such as PIXART-alpha, PIXART-Sigma, and the 4K-capable Wuerstchen pipeline. In parallel, flow matching techniques inspired by Rectified Flow learn a more direct trajectory from noise to image, enabling remarkably fast sampling. InstaFlow showed that a single denoising step could suffice for high-quality text-to-image synthesis, and the state-of-the-art SD3 model combines both trends by pairing rectified flow with transformer backbones. The latest generation, exemplified by FLUX.1 Kontext, unifies image generation and editing within a single flow matching architecture, maintaining multi-turn consistency as users iteratively refine their creations.

Video generation has followed a parallel but more demanding trajectory, because adding a temporal dimension multiplies computational complexity and introduces the problem of temporal coherence. Early systems such as Video Diffusion Models extended 2D U-Nets to 3D by inserting temporal attention layers, while Make-A-Video, Imagen Video, and MagicVideo demonstrated that large-scale image priors could be transferred to video with relatively little paired text-video data. Latent-space approaches like LVDM compressed video into low-dimensional representations for efficient long-form generation, and Stable Video Diffusion scaled latent video diffusion to large datasets by building on pretrained image models. More recently, transformer-based architectures including CogVideoX, LATTE, and OpenAI’s Sora have adopted the Diffusion Transformer to better fuse linguistic semantics with long-range motion, while industrial-scale systems such as HunyuanVideo, Kling, Mochi, Wan, and SkyReels-V1 continue to push quality and generalization boundaries.

Yet text prompts alone proved too blunt an instrument for creators who need precise control over spatial layout, human pose, or camera motion. The survey highlights how researchers answered with structured conditioning. ControlNet, perhaps the most influential contribution, attaches lightweight zero-initialized branches to a frozen diffusion model, translating depth maps, edges, poses, and segmentation masks into feature-space guidance without disturbing the base model’s knowledge. Video systems like Gen-1 decoupled structure from content, using monocular depth as a motion scaffold and CLIP embeddings as appearance signals. A second wave of 3D-aware methods goes further, injecting camera trajectories as ray embeddings or generating coarse point clouds as geometric scaffolds before refinement. Systems such as CameraCtrl, ViewCrafter, GEN3C, and Uni3C enable explicit viewpoint control and novel view synthesis, opening applications in filmmaking, simulation, and robotics.

What makes the survey striking is its documentation of the reverse direction: diffusion models being turned into perception engines. Rather than classifying or regressing, generative approaches now treat understanding tasks as conditional generation problems. Marigold reframes monocular depth estimation as a latent diffusion task and achieves strong zero-shot performance by repurposing pretrained image generators. DepthFM accelerates the idea with flow matching, while StableNormal and NormalCrafter generate temporally consistent surface normal maps from images and video. Segmentation has followed suit: DiffusionInst and ODISE formulate instance and panoptic segmentation as conditional generation, refining masks through denoising, and training-free methods show that powerful semantic cues are already embedded in pretrained diffusion backbones, extractable through feature clustering and attention analysis without any task-specific training. Joint frameworks like GeoWizard, Lotus, GeometryCrafter, Geo4D, and DICEPTION extend this to simultaneous prediction of depth, normals, camera parameters, and point clouds, often surpassing discriminative specialists in zero-shot and cross-domain settings, though the multi-step denoising process still imposes a significant inference cost.

The survey also maps how models are learning to generate multiple modalities at once, moving beyond RGB toward outputs that carry explicit geometry. Two architectural strategies dominate. Multi-modal VAE approaches, exemplified by Orchid and Trellis, encode RGB, depth, and normals, or point clouds and semantic features, into a shared latent space decoded into different modality-specific outputs, enabling coherent joint synthesis for 3D asset creation and sensor fusion. Shared VAE approaches instead pass all modalities through a single autoencoder: Matrix3D performs pose estimation, depth prediction, and novel view synthesis in one masked multi-modal diffusion transformer; VideoJAM jointly models appearance and optical flow with an inner-guidance mechanism for temporally coherent motion; Voyager synthesizes aligned RGB and depth video for unbounded, world-consistent 3D scene expansion; and JointDiT captures the joint RGB-depth distribution through adaptive weighting. Together they signal a shift toward flexible latent frameworks that can mask, fuse, and weigh heterogeneous signals spanning appearance, geometry, semantics, and motion.

At the frontier lies the boldest ambition: unified models that generate and understand in a single system. The authors trace a lineage from masked-autoencoder and tokenized approaches such as MultiMAE, UNIFIED-IO and its sequel, the 4M family, Sapiens, and PixelWorld, which consolidate dozens of tasks and modalities into shared token spaces. Diffusion-based successors now close the loop. UniReal treats image tasks as discontinuous video frames, OneDiff frames tasks as views at different noise levels, MMGen and OmniGen2 unify generation, editing, and understanding with parallel transformers, and on the video side OmniVDiff jointly models RGB, depth, segmentation, and edges through a shared 3D VAE, while Aether demonstrates zero-shot 4D reconstruction and goal-driven planning, and VACE consolidates creation and editing operations from move-anything to animate-anything in one latent diffusion framework. Applications already span world models for autonomous driving such as STAG-1 and Cosmos-Transfer1, immersive 3D and 4D scene generation via Matrix3D and Geo4D, and flexible video editing pipelines.

Challenges remain substantial, the survey cautions. How modalities interact during joint training is poorly understood, with some signals helping and others injecting noise or redundancy. Unified models still trail unimodal specialists on per-modality accuracy, slow multi-step inference limits deployment, and long-range video with temporally consistent multi-modal coherence remains unsolved. Bridging synthetic generation with real-world interaction, essential for embodied AI and robotics, is the ultimate test. Still, the trajectory is unmistakable: diffusion models have evolved from novelty image generators into a general substrate for visual intelligence, and the survey’s vision of systems that simultaneously perceive, reason about, and imagine the world is no longer science fiction but an active engineering frontier, one whose benchmarks, efficiency innovations, and unified architectures will shape the next decade of computer vision research.

Subject of Research: A survey: advances in multi-modal visual understanding and generation

Article Title: A survey: advances in multi-modal visual understanding and generation

Article References: Xi, D., Zhou, Z., Wang, Y., Wang, J., & Zhang, C. (2026). A survey: advances in multi-modal visual understanding and generation. Vicinagearth, 3(1), Article 8. https://doi.org/10.1007/s44336-025-00032-x

Image Credits: AI Generated

DOI: 10.1007/s44336-025-00032-x

Keywords: survey, advances, multi-modal, visual, understanding, generation, scientific research

Centromeric chromatin is uniquely organized to balance structural rigidity with dynamic functional requirements. Central to this organization is the enrichment of nucleosomes containing CENP-A, which demarcates the core centromeric domains. Flanking these domains are pericentromeric regions enriched with repressive heterochromatin marks, notably di- and tri-methylation of lysine 9 on histone H3 (H3K9me2 and H3K9me3). Historically, the trimethylated form, H3K9me3, deposited by the SUV39H methyltransferase, has been recognized as a canonical marker of transcriptionally silent heterochromatin, critical for maintaining chromosomal stability. However, the biological significance and mechanistic roles of G9a (also known as EHMT2), the primary methyltransferase responsible for generating H3K9me2 at the (peri)centromeric regions, have remained enigmatic until this recent investigation.

The researchers harnessed a multifaceted experimental approach combining chromatin immunoprecipitation, advanced imaging, and functional genomics to elucidate how G9a-mediated H3K9me2 influences centromeric chromatin configuration and mitotic fidelity. Key findings revealed that G9a/GLP methyltransferase complexes localize specifically to centromeres, enriching local H3K9me2 without significantly affecting adjacent heterochromatin domains. This localized modification is indispensable for the orchestration of a chromatin environment conducive to accurate chromosome segregation. Notably, depletion of G9a leads to a pronounced prolongation of mitosis and an increase in chromosomal missegregations, hallmark indicators of compromised centromere function.

Intriguingly, the study uncovered that G9a-mediated H3K9me2 acts as a gatekeeper that limits the spatial distribution of the repressive H3K9me3 mark, effectively preventing its encroachment into the core centromeric domains. This containment preserves the integrity of centromeric chromatin and facilitates the recruitment and efficient functioning of RNA Polymerase II locally. Consequent transcriptional activity at the centromere gives rise to R-loops, three-stranded nucleic acid structures comprised of an RNA-DNA hybrid and displaced single-stranded DNA, which are increasingly recognized as regulatory elements in chromatin biology. The maintenance of these R-loops is essential for proper loading of Replication Protein A (RPA), a single-strand DNA-binding protein critical for DNA replication and repair processes, including activation of the ATR-Chk1 DNA damage response pathway.

G9a knockdown elicited a cascade of epigenetic and transcriptional disruptions, starting with the invasion of H3K9me3 into centromeric cores, leading to diminished recruitment of RNA Polymerase II. This diminished transcriptional output compromised the formation of centromeric R-loops, thereby impairing RPA loading. The attenuation of RPA occupancy undermines the activation of the ATR-Chk1-Aurora B kinase signaling axis, a crucial cascade that ensures correction of erroneous kinetochore-microtubule attachments during mitosis. Aurora B kinase functions as a central mitotic regulator to resolve attachment errors, thus ensuring chromosome biorientation and faithful segregation. Loss of G9a activity therefore precipitates a failure in this checkpoint mechanism, resulting in chromosome missegregations and subsequent genomic instability.

Furthermore, the study highlighted alterations in the centromeric deposition of CENP-A upon G9a depletion. As CENP-A nucleosomes are fundamental for kinetochore nucleation and function, their misincorporation exacerbates centromere dysfunction, further contributing to chromosomal instability. These findings provide compelling evidence linking epigenetic regulation by G9a to the structural and functional stability of centromeres. The link between G9a somatic mutations and defective chromosome segregation observed in cancer cells offers a molecular explanation for the tumorigenic potential associated with such mutations. This insight underscores the clinical relevance of G9a-mediated epigenetic modulation in the pathogenesis of genomic instability-related diseases, including cancer.

Mechanistically, the data posit that G9a-mediated H3K9me2 establishes a unique chromatin landscape that integrates transcriptional activity, DNA replication stress responses, and mitotic checkpoint signaling. By preventing the unwarranted spread of heterochromatin marks and promoting centromeric transcription coupled with R-loop formation, G9a ensures the proper orchestration of protein complexes essential for checkpoint surveillance and chromosome segregation. This multifaceted control reveals a novel layer of epigenetic regulation that bridges chromatin dynamics with cell cycle progression and genome maintenance.

From a therapeutic standpoint, these findings open promising new avenues for targeting G9a and its associated pathways in diseases characterized by chromosomal instability, most notably cancer. Small molecule inhibitors of G9a have been explored primarily in the context of their influence on global gene expression and oncogenic pathways; however, their impact on centromere integrity and mitotic fidelity warrants careful reevaluation. Therapeutic modulation of G9a activity could potentially restore proper centromere function or selectively sensitize tumor cells exhibiting G9a dysfunction, highlighting the need for comprehensive studies assessing G9a-targeted interventions within the framework of chromosomal stability.

In conclusion, this landmark study significantly advances our understanding of the epigenetic regulation governing centromere integrity. It establishes G9a-mediated H3K9me2 as a crucial safeguard that choreographs a delicate balance between heterochromatin demarcation, centromeric transcription, R-loop dynamics, and mitotic checkpoint activation. The elucidation of G9a’s role in maintaining chromosome segregation fidelity not only deepens fundamental chromatin biology but also has profound implications for therapeutic strategies aimed at combating genome instability in human disease. As research continues to unravel the complexity of centromeric epigenetics, these insights position G9a as a central node connecting chromatin modification enzymes with the maintenance of genome integrity during cell division.

Subject of Research: Epigenetic regulation of centromere integrity and chromosome segregation

Article Title: G9a-mediated H3K9me2 preserves centromere integrity and facilitates faithful chromosome segregation

News Publication Date: Not specified

Web References: http://dx.doi.org/10.1016/j.scib.2025.08.023

Image Credits: ©Science China Press

Keywords: G9a, H3K9me2, centromere integrity, chromosome segregation, CENP-A, heterochromatin, R-loops, ATR-Chk1-Aurora B pathway, epigenetics, mitosis, genome stability, chromosome missegregation