Omni / 多模态生成调研 — 全量来源索引

全量来源索引（按发布年月）

去重后 344 条一手来源（按 date 年份归组；去重键为原始 URL）。

2020（7 条）

• Jukebox: A Generative Model for Music — OpenAI · 2020-04 · paper · [audio] — https://openai.com/index/jukebox/
• Denoising Diffusion Probabilistic Models (DDPM) — UC Berkeley · 2020-06 · paper · [method] — https://arxiv.org/abs/2006.11239
• iGPT) — OpenAI · 2020-06 · paper · [method] — https://openai.com/index/image-gpt/
• Improved Techniques for Training Score-Based Generative Models (NCSNv2) — Stanford University · 2020-06 · paper · [method] — https://arxiv.org/abs/2006.09011
• Denoising Diffusion Implicit Models (DDIM) — Stanford University · 2020-10 · paper · [method] — https://arxiv.org/abs/2010.02502
• Score-Based Generative Modeling through Stochastic Differential Equations (Score SDE) — Stanford University / Google Brain · 2020-11 · paper · [method] — https://openreview.net/forum?id=PxTIG12RRHS
• Taming Transformers for High-Resolution Image Synthesis (VQGAN) — Heidelberg University (CompVis) · 2020-12 · paper · [method] — https://arxiv.org/abs/2012.09841

2021（25 条）

• DALL·E: Zero-Shot Text-to-Image Generation — OpenAI · 2021-02 · paper · [t2i] — https://openai.com/index/dall-e/
• IDDPM) — OpenAI · 2021-02 · paper · [method] — https://arxiv.org/abs/2102.09672
• M6: A Chinese Multimodal Pretrainer — Alibaba DAMO Academy · 2021-03 · paper · [unified] — https://arxiv.org/abs/2103.00823
• Paint by Word — MIT CSAIL / Harvard GSD · 2021-03 · paper · [edit] — https://arxiv.org/abs/2103.10951
• StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery — Hebrew University of Jerusalem / Tel-Aviv University / Adobe Research · 2021-03 · paper · [edit] — https://arxiv.org/abs/2103.17249
• GODIVA: Generating Open-DomaIn Videos from nAtural Descriptions — Microsoft Research Asia / Duke University · 2021-04 · paper · [video] — https://arxiv.org/abs/2104.14806
• VQGAN-CLIP: Open Domain Image Generation and Editing with Natural Language Guidance — EleutherAI / community (Crowson et al.) · 2021-04 · paper · [t2i] — https://arxiv.org/abs/2204.08583
• VideoGPT: Video Generation using VQ-VAE and Transformers — UC Berkeley · 2021-04 · paper · [video] — https://arxiv.org/abs/2104.10157
• 社区版) — Community (Katherine Crowson, RiversHaveWings) / EleutherAI · 2021-05 · method · [method] — https://colab.research.google.com/drive/1QBsaDAZv8np29FPbvjffbE1eytoJcsgA
• Cascaded Diffusion Models for High Fidelity Image Generation (CDM) — Google Research / Brain · 2021-05 · paper · [method] — https://arxiv.org/abs/2106.15282
• CogView: Mastering Text-to-Image Generation via Transformers — Tsinghua / DAMO Academy (Alibaba) / BAAI · 2021-05 · paper · [t2i] — https://arxiv.org/abs/2105.13290
• Classifier Guidance) — OpenAI · 2021-05 · paper · [method] — https://arxiv.org/abs/2105.05233
• ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models — Seoul National University / Samsung SDS · 2021-08 · paper · [method] — https://arxiv.org/abs/2108.02938
• SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations — Stanford University / Carnegie Mellon University · 2021-08 · paper · [edit] — https://arxiv.org/abs/2108.01073
• CLIP-Forge: Towards Zero-Shot Text-to-Shape Generation — Autodesk AI Lab / Autodesk Research · 2021-10 · paper · [3d] — https://arxiv.org/abs/2110.02624
• Vector-quantized Image Modeling with Improved VQGAN (ViT-VQGAN) — Google Research · 2021-10 · paper · [method] — https://arxiv.org/abs/2110.04627
• Blended Diffusion for Text-driven Editing of Natural Images — The Hebrew University of Jerusalem & Reichman University · 2021-11 · paper · [edit] — https://arxiv.org/abs/2111.14818
• PoE-GAN — 文字 + 语义涂鸦 + 草图多模态合成风景图 — NVIDIA · 2021-11 · blog · [t2i] — https://blogs.nvidia.com/blog/2021/11/22/gaugan2-ai-art-demo/
• NÜWA: Visual Synthesis Pre-training for Neural visUal World creAtion — Microsoft Research Asia / Peking University · 2021-11 · paper · [unified] — https://arxiv.org/abs/2111.12417
• Palette: Image-to-Image Diffusion Models — Google Research (Brain Team) · 2021-11 · paper · [edit] — https://arxiv.org/abs/2111.05826
• VQ-Diffusion：用于文生图的向量量化扩散模型 — Microsoft Research / 中国科学技术大学 (USTC) · 2021-11 · paper · [method] — https://arxiv.org/abs/2111.14822
• ERNIE-ViLG: Unified Generative Pre-training for Bidirectional Vision-Language Generation — Baidu · 2021-12 · paper · [unified] — https://arxiv.org/abs/2112.15283
• GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models — OpenAI · 2021-12 · paper · [t2i] — https://arxiv.org/abs/2112.10741
• High-Resolution Image Synthesis with Latent Diffusion Models (LDM) — LMU Munich / IWR Heidelberg (CompVis) / Runway ML · 2021-12 · paper · [foundation] — https://arxiv.org/abs/2112.10752
• Zero-Shot Text-Guided Object Generation with Dream Fields — Google Research / UC Berkeley · 2021-12 · paper · [3d] — https://ajayj.com/dreamfields

2022（44 条）

• CM3: A Causal Masked Multimodal Model of the Internet — Meta (FAIR) · 2022-01 · paper · [unified] — https://arxiv.org/abs/2201.07520
• RePaint: Inpainting using Denoising Diffusion Probabilistic Models — ETH Zürich (Computer Vision Lab) · 2022-01 · paper · [edit] — https://arxiv.org/abs/2201.09865
• MaskGIT: Masked Generative Image Transformer — Google Research · 2022-02 · paper · [method] — https://arxiv.org/abs/2202.04200
• StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets — University of Tübingen / Max Planck Institute for Intelligent Systems · 2022-02 · paper · [t2i] — https://arxiv.org/abs/2202.00273
• Make-A-Scene: Scene-Based Text-to-Image Generation with Human Priors — Meta AI Research · 2022-03 · paper · [t2i] — https://arxiv.org/abs/2203.13131
• CogView2: Faster and Better Text-to-Image Generation via Hierarchical Transformers — Tsinghua / BAAI · 2022-04 · paper · [t2i] — https://arxiv.org/abs/2204.14217
• unCLIP（Hierarchical Text-Conditional Image Generation with CLIP Latents） — OpenAI · 2022-04 · paper · [t2i] — https://arxiv.org/abs/2204.06125
• Video Diffusion Models (VDM) — Google Research · 2022-04 · paper · [video] — https://arxiv.org/abs/2204.03458
• CogVideo: Large-scale Pretraining for Text-to-Video Generation via Transformers — Tsinghua University / BAAI · 2022-05 · paper · [video] — https://arxiv.org/abs/2205.15868
• Imagen: Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding — Google Research (Brain Team) · 2022-05 · paper · [t2i] — https://arxiv.org/abs/2205.11487
• Blended Latent Diffusion — The Hebrew University of Jerusalem / Reichman University · 2022-06 · paper · [edit] — https://arxiv.org/abs/2206.02779
• DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps — Tsinghua University (TSAIL) & Renmin University · 2022-06 · paper · [method] — https://arxiv.org/abs/2206.00927
• Elucidating the Design Space of Diffusion-Based Generative Models (EDM) — NVIDIA · 2022-06 · paper · [method] — https://arxiv.org/abs/2206.00364
• Parti: Scaling Autoregressive Models for Content-Rich Text-to-Image Generation — Google Research · 2022-06 · paper · [t2i] — https://arxiv.org/abs/2206.10789
• Classifier-Free Diffusion Guidance (CFG) — Google Research, Brain team · 2022-07 · paper · [method] — https://arxiv.org/abs/2207.12598
• Midjourney (公测 V1–V4) — Midjourney · 2022-07 · blog · [t2i] — https://docs.midjourney.com/hc/en-us/articles/33329788681101-Legacy-Features
• An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion — Tel Aviv University / NVIDIA · 2022-08 · paper · [edit] — https://arxiv.org/abs/2208.01618
• DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation — Google Research / Boston University · 2022-08 · paper · [edit] — https://arxiv.org/abs/2208.12242
• Prompt-to-Prompt Image Editing with Cross Attention Control — Google Research / Tel Aviv University · 2022-08 · paper · [edit] — https://arxiv.org/abs/2208.01626
• Runway) — Stability AI / CompVis / Runway ML · 2022-08 · model-card · [t2i] — https://github.com/CompVis/stable-diffusion
• All are Worth Words: A ViT Backbone for Diffusion Models (U-ViT) — Tsinghua University (TSAIL) / Renmin University of China / BAAI · 2022-09 · paper · [method] — https://arxiv.org/abs/2209.12152
• AudioLM: a Language Modeling Approach to Audio Generation — Google Research · 2022-09 · paper · [audio] — https://arxiv.org/abs/2209.03143
• DreamFusion: Text-to-3D using 2D Diffusion — Google Research / UC Berkeley · 2022-09 · paper · [3d] — https://arxiv.org/abs/2209.14988
• Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow — UT Austin · 2022-09 · paper · [method] — https://arxiv.org/abs/2209.03003
• GET3D: A Generative Model of High Quality 3D Textured Shapes Learned from Images — NVIDIA · 2022-09 · paper · [3d] — https://nv-tlabs.github.io/GET3D/
• Make-A-Video: Text-to-Video Generation without Text-Video Data — Meta AI · 2022-09 · paper · [video] — https://arxiv.org/abs/2209.14792
• DiffEdit: Diffusion-based Semantic Image Editing with Mask Guidance — Meta AI / Sorbonne Université / Valeo.ai · 2022-10 · paper · [edit] — https://arxiv.org/abs/2210.11427
• ERNIE-ViLG 2.0: Improving Text-to-Image Diffusion Model with Knowledge-Enhanced Mixture-of-Denoising-Experts — Baidu · 2022-10 · paper · [t2i] — https://arxiv.org/abs/2210.15257
• Flow Matching for Generative Modeling — Meta AI (FAIR) / Weizmann Institute · 2022-10 · paper · [method] — https://arxiv.org/abs/2210.02747
• Imagen Video: High Definition Video Generation with Diffusion Models — Google Research (Brain Team) · 2022-10 · tech-report · [video] — https://imagen.research.google/video/
• Imagic: Text-Based Real Image Editing with Diffusion Models — Google Research / Technion / Weizmann Institute · 2022-10 · paper · [edit] — https://arxiv.org/abs/2210.09276
• Phenaki: Variable Length Video Generation from Open Domain Textual Descriptions — Google Brain · 2022-10 · paper · [video] — https://openreview.net/forum?id=vOEXS39nOF
• img2img (SD v1.5 编辑权重) — Runway / Stability AI / CompVis · 2022-10 · model-card · [edit] — https://huggingface.co/runwayml/stable-diffusion-inpainting
• InstructPix2Pix: Learning to Follow Image Editing Instructions — UC Berkeley · 2022-11 · paper · [edit] — https://arxiv.org/abs/2211.09800
• Magic3D: High-Resolution Text-to-3D Content Creation — NVIDIA · 2022-11 · paper · [3d] — https://research.nvidia.com/labs/dir/magic3d/
• Null-text Inversion for Editing Real Images using Guided Diffusion Models — Google Research / Tel Aviv University · 2022-11 · paper · [edit] — https://arxiv.org/abs/2211.09794
• 2.1 — Stability AI · 2022-11 · model-card · [t2i] — https://stability.ai/news/stable-diffusion-v2-release
• Versatile Diffusion: Text, Images and Variations All in One Diffusion Model — SHI Labs (UIUC / Georgia Tech / U of Oregon) · UT Austin · Picsart AI Research (PAIR) · 2022-11 · paper · [unified] — https://arxiv.org/abs/2211.08332
• eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers — NVIDIA · 2022-11 · tech-report · [t2i] — https://arxiv.org/abs/2211.01324
• MAGVIT: Masked Generative Video Transformer — Google Research / Carnegie Mellon University · 2022-12 · paper · [video] — https://arxiv.org/abs/2212.05199
• Point·E: A System for Generating 3D Point Clouds from Complex Prompts — OpenAI · 2022-12 · paper · [3d] — https://arxiv.org/abs/2212.08751
• Scalable Diffusion Models with Transformers (DiT) — UC Berkeley / New York University (Meta AI FAIR) · 2022-12 · paper · [method] — https://arxiv.org/abs/2212.09748
• Tune-A-Video: One-Shot Tuning of Image Diffusion Models for Text-to-Video Generation — Show Lab (NUS) / Tencent ARC Lab & PCG · 2022-12 · paper · [video] — https://arxiv.org/abs/2212.11565
• UniDiffuser: One Transformer Fits All Distributions in Multi-Modal Diffusion — Tsinghua (TSAIL) / RUC / ShengShu / BAAI · 2022-12 · paper · [unified] — https://arxiv.org/abs/2303.06555

2023（92 条）

• GLIGEN: Open-Set Grounded Text-to-Image Generation — UW-Madison / Columbia / Microsoft · 2023-01 · paper · [edit] — https://arxiv.org/abs/2301.07093
• Muse: Text-To-Image Generation via Masked Generative Transformers — Google Research · 2023-01 · paper · [t2i] — https://arxiv.org/abs/2301.00704
• MusicLM: Generating Music From Text — Google Research · 2023-01 · paper · [audio] — https://arxiv.org/abs/2301.11325
• StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis — NVIDIA / University of Tübingen · 2023-01 · paper · [t2i] — https://arxiv.org/abs/2301.09515
• VALL-E: Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers — Microsoft · 2023-01 · paper · [audio] — https://arxiv.org/abs/2301.02111
• ControlNet: Adding Conditional Control to Text-to-Image Diffusion Models — Stanford University · 2023-02 · paper · [edit] — https://arxiv.org/abs/2302.05543
• Gen-1: Structure and Content-Guided Video Synthesis with Diffusion Models — Runway · 2023-02 · paper · [video] — https://arxiv.org/abs/2302.03011
• Pix2Pix-Zero: Zero-shot Image-to-Image Translation — CMU / Adobe Research · 2023-02 · paper · [edit] — https://arxiv.org/abs/2302.03027
• T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models — Tencent ARC Lab (PCG) / 北京大学 · 2023-02 · paper · [edit] — https://arxiv.org/abs/2302.08453
• Adobe Firefly（首个图像模型 + Photoshop Generative Fill） — Adobe · 2023-03 · blog · [t2i] — https://news.adobe.com/news/news-details/2023/adobe-unveils-firefly-a-family-of-new-creative-generative-ai
• Consistency Models — OpenAI · 2023-03 · paper · [method] — https://arxiv.org/abs/2303.01469
• Zero-1-to-3: Zero-shot One Image to 3D Object — Columbia University / Toyota Research Institute · 2023-03 · paper · [3d] — https://arxiv.org/abs/2303.11328
• Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models (Video LDM) — NVIDIA / LMU Munich · 2023-04 · paper · [video] — https://arxiv.org/abs/2304.08818
• DeepFloyd IF — Stability AI / DeepFloyd · 2023-04 · model-card · [t2i] — https://github.com/deep-floyd/IF
• ImageReward & ReFL: Learning and Evaluating Human Preferences for Text-to-Image Generation — Tsinghua University / Zhipu AI · 2023-04 · paper · [method] — https://arxiv.org/abs/2304.05977
• InstantBooth: Personalized Text-to-Image Generation without Test-Time Finetuning — Adobe Inc. · 2023-04 · paper · [edit] — https://arxiv.org/abs/2304.03411
• MasaCtrl: Tuning-Free Mutual Self-Attention Control for Consistent Image Synthesis and Editing — The University of Tokyo / Tencent PCG (ARC Lab) · 2023-04 · paper · [edit] — https://arxiv.org/abs/2304.08465
• DDPO: Training Diffusion Models with Reinforcement Learning — UC Berkeley / MIT · 2023-05 · paper · [method] — https://arxiv.org/abs/2305.13301
• DragDiffusion: 基于点的交互式拖拽图像编辑 — MPI-Informatics / NUS / ByteDance · 2023-05 · paper · [edit] — https://arxiv.org/abs/2305.10973
• GILL: Generating Images with Multimodal Language Models — Carnegie Mellon University (CMU) · 2023-05 · paper · [unified] — https://arxiv.org/abs/2305.17216
• ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillation — Tsinghua University / Renmin University / ShengShu · 2023-05 · paper · [3d] — https://ml.cs.tsinghua.edu.cn/prolificdreamer/
• RAPHAEL: Text-to-Image Generation via Large Mixture of Diffusion Paths — SenseTime Research / HKU · 2023-05 · paper · [t2i] — https://arxiv.org/abs/2305.18295
• Uni-ControlNet: All-in-One Control to Text-to-Image Diffusion Models — HKU / Microsoft · 2023-05 · paper · [edit] — https://arxiv.org/abs/2305.16322
• UniControl: A Unified Diffusion Model for Controllable Visual Generation In the Wild — Salesforce AI Research / Northeastern / Stanford · 2023-05 · paper · [edit] — https://arxiv.org/abs/2305.11147
• COFT) — MPI for Intelligent Systems · Cambridge · Tübingen · 2023-06 · paper · [method] — https://arxiv.org/abs/2306.07280
• Human Preference Score v2 (HPS v2): A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis — CUHK MMLab / CPII / SenseTime · 2023-06 · paper · [method] — https://arxiv.org/abs/2306.09341
• MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing — Ohio State University / University of Waterloo · 2023-06 · paper · [edit] — https://osu-nlp-group.github.io/MagicBrush/
• AudioCraft: Simple and Controllable Music Generation — Meta AI · 2023-06 · paper · [audio] — https://arxiv.org/abs/2306.05284
• One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimization — UC San Diego / UCLA / Cornell / 浙大 / IIT Madras / Adobe · 2023-06 · paper · [3d] — https://arxiv.org/abs/2306.16928
• Rerender A Video: Zero-Shot Text-Guided Video-to-Video Translation — NTU S-Lab · 2023-06 · paper · [video] — https://arxiv.org/abs/2306.07954
• StyleDrop: Text-to-Image Generation in Any Style — Google Research · 2023-06 · paper · [edit] — https://arxiv.org/abs/2306.00983
• Würstchen: An Efficient Architecture for Large-Scale Text-to-Image Diffusion Models — Wand Technologies / TH Ingolstadt / Mila / Stability AI(compute) · 2023-06 · paper · [t2i] — https://arxiv.org/abs/2306.00637
• AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning — CUHK / Shanghai AI Lab / Stanford · 2023-07 · paper · [video] — https://arxiv.org/abs/2307.04725
• CM3Leon: Scaling Autoregressive Multi-Modal Models (Pretraining and Instruction Tuning) — Meta FAIR · 2023-07 · paper · [unified] — https://arxiv.org/abs/2309.02591
• Emu: Generative Pretraining in Multimodality — BAAI (智源) + 清华 THU + 北大 PKU · 2023-07 · paper · [unified] — https://arxiv.org/abs/2307.05222
• HyperDreamBooth: HyperNetworks for Fast Personalization of Text-to-Image Models — Google Research · 2023-07 · paper · [edit] — https://arxiv.org/abs/2307.06949
• Kandinsky 2.x (Image Prior + Latent Diffusion) — Sber AI / AIRI / Skoltech · 2023-07 · tech-report · [t2i] — https://arxiv.org/abs/2310.03502
• SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis — Stability AI · 2023-07 · tech-report · [t2i] — https://arxiv.org/abs/2307.01952
• SEED: Planting a SEED of Vision in Large Language Model — Tencent AI Lab / ARC Lab (Tencent PCG) · 2023-07 · paper · [unified] — https://arxiv.org/abs/2307.08041
• Subject-Diffusion: Open Domain Personalized Text-to-Image Generation without Test-time Fine-tuning — OPPO AI Center · 2023-07 · paper · [edit] — https://arxiv.org/abs/2307.11410
• TokenFlow: Consistent Diffusion Features for Consistent Video Editing — Weizmann Institute of Science · 2023-07 · paper · [video] — https://arxiv.org/abs/2307.10373
• 3D Gaussian Splatting for Real-Time Radiance Field Rendering — Inria / Université Côte d’Azur (GRAPHDECO) + MPI Informatik · 2023-08 · paper · [3d] — https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/
• AudioLDM 2: Learning Holistic Audio Generation with Self-supervised Pretraining — University of Surrey (CVSSP) / CUHK / ByteDance SAMI · 2023-08 · paper · [audio] — https://arxiv.org/abs/2308.05734
• IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models — Tencent AI Lab · 2023-08 · paper · [edit] — https://arxiv.org/abs/2308.06721
• Ideogram 0.1 — Ideogram AI · 2023-08 · blog · [t2i] — https://ideogram.ai/launch
• MVDream: Multi-view Diffusion for 3D Generation — ByteDance · 2023-08 · paper · [3d] — https://arxiv.org/abs/2308.16512
• ModelScopeT2V（ModelScope Text-to-Video Technical Report） — Alibaba Group / DAMO Academy · 2023-08 · tech-report · [video] — https://arxiv.org/abs/2308.06571
• AnyMAL: An Efficient and Scalable Any-Modality Augmented Language Model — Meta (FAIR & Reality Labs) · 2023-09 · paper · [omni] — https://arxiv.org/abs/2309.16058
• DreamLLM: Synergistic Multimodal Comprehension and Creation — Xi’an Jiaotong University / MEGVII / Tsinghua · 2023-09 · paper · [unified] — https://openreview.net/forum?id=y01KGvd9Bw
• Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack — Meta · 2023-09 · paper · [t2i] — https://arxiv.org/abs/2309.15807
• FreeU: Free Lunch in Diffusion U-Net — S-Lab, Nanyang Technological University · 2023-09 · paper · [method] — https://arxiv.org/abs/2309.11497
• InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation — UT Austin / Helixon Research · 2023-09 · paper · [method] — https://arxiv.org/abs/2309.06380
• InstructDiffusion: A Generalist Modeling Interface for Vision Tasks — Microsoft Research Asia · 2023-09 · paper · [edit] — https://arxiv.org/abs/2309.03895
• LaVie: High-Quality Video Generation with Cascaded Latent Diffusion Models — Shanghai AI Lab · 2023-09 · paper · [video] — https://arxiv.org/abs/2309.15103
• MGIE: Guiding Instruction-based Image Editing via Multimodal Large Language Models — Apple / UC Santa Barbara · 2023-09 · paper · [edit] — https://arxiv.org/abs/2309.17102
• NExT-GPT: Any-to-Any Multimodal LLM — National University of Singapore (NExT++) · 2023-09 · paper · [omni] — https://arxiv.org/abs/2309.05519
• PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis — Huawei Noah’s Ark Lab · 2023-09 · paper · [t2i] — https://arxiv.org/abs/2310.00426
• Show-1: Marrying Pixel and Latent Diffusion Models for Text-to-Video Generation — Show Lab, National University of Singapore · 2023-09 · paper · [video] — https://arxiv.org/abs/2309.15818
• SyncDreamer: Generating Multiview-consistent Images from a Single-view Image — HKU / Tencent Games / UPenn / Texas A&M · 2023-09 · paper · [3d] — https://arxiv.org/abs/2309.03453
• CommonCanvas: An Open Diffusion Model Trained with Creative-Commons Images — Cornell Tech & Databricks Mosaic (MosaicML) · 2023-10 · paper · [t2i] — https://arxiv.org/abs/2310.16825
• DALL·E 3 — OpenAI · 2023-10 · blog · [t2i] — https://openai.com/index/dall-e-3/
• HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion — Snap Inc. / CUHK / HKU / NTU · 2023-10 · paper · [t2i] — https://arxiv.org/abs/2310.08579
• Kosmos-G: Generating Images in Context with Multimodal Large Language Models — Microsoft Research · 2023-10 · paper · [unified] — https://arxiv.org/abs/2310.02992
• LCM-LoRA) — Tsinghua IIIS (+ Hugging Face for LCM-LoRA) · 2023-10 · paper · [method] — https://arxiv.org/abs/2310.04378
• MAGVIT-v2 — Language Model Beats Diffusion: Tokenizer is Key to Visual Generation — Google / Carnegie Mellon University · 2023-10 · paper · [method] — https://arxiv.org/abs/2310.05737
• VideoCrafter1: Open Diffusion Models for High-Quality Video Generation — Tencent AI Lab · 2023-10 · tech-report · [video] — https://arxiv.org/abs/2310.19512
• Wonder3D: Single Image to 3D using Cross-Domain Diffusion — HKU / VAST / Tsinghua · 2023-10 · paper · [3d] — https://arxiv.org/abs/2310.15008
• Zero123++: a Single Image to Consistent Multi-view Diffusion Base Model — SUDO-AI (UC San Diego / Stanford / Tsinghua / UCLA / Zhejiang Univ.) · 2023-10 · tech-report · [3d] — https://arxiv.org/abs/2310.15110
• Animate Anyone: Consistent and Controllable Image-to-Video Synthesis for Character Animation — Alibaba (Institute for Intelligent Computing) · 2023-11 · paper · [video] — https://arxiv.org/abs/2311.17117
• Concept Sliders: LoRA Adaptors for Precise Control in Diffusion Models — Northeastern University / MIT · 2023-11 · paper · [edit] — https://arxiv.org/abs/2311.12092
• D3PO: Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model — Tsinghua University (SIGS) / Parametrix Technology · 2023-11 · paper · [method] — https://arxiv.org/abs/2311.13231
• Diffusion-DPO: Diffusion Model Alignment Using Direct Preference Optimization — Salesforce AI / Stanford University · 2023-11 · paper · [method] — https://arxiv.org/abs/2311.12908
• Emu Edit: Precise Image Editing via Recognition and Generation Tasks — Meta (GenAI / AI at Meta) · 2023-11 · paper · [edit] — https://emu-edit.metademolab.com/
• Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioning — Meta · 2023-11 · paper · [video] — https://emu-video.metademolab.com/
• I2VGen-XL: High-Quality Image-to-Video Synthesis via Cascaded Diffusion Models — Alibaba Group (Tongyi Lab) · 2023-11 · tech-report · [video] — https://arxiv.org/abs/2311.04145
• LCM-LoRA: A Universal Stable-Diffusion Acceleration Module — Tsinghua University (IIIS) / Hugging Face · 2023-11 · tech-report · [method] — https://arxiv.org/abs/2311.05556
• LEDITS++: Limitless Image Editing using Text-to-Image Models — TU Darmstadt / DFKI / hessian.AI / Hugging Face · 2023-11 · paper · [edit] — https://arxiv.org/abs/2311.16711
• LRM: Large Reconstruction Model for Single Image to 3D — Adobe Research / Australian National University · 2023-11 · paper · [3d] — https://arxiv.org/abs/2311.04400
• MagicAnimate: Temporally Consistent Human Image Animation using Diffusion Model — Show Lab (NUS) / ByteDance · 2023-11 · paper · [video] — https://arxiv.org/abs/2311.16498
• Make Pixels Dance: High-Dynamic Video Generation (PixelDance) — ByteDance Research · 2023-11 · paper · [video] — https://arxiv.org/abs/2311.10982
• One-step Diffusion with Distribution Matching Distillation (DMD) — MIT / Adobe Research · 2023-11 · paper · [method] — https://arxiv.org/abs/2311.18828
• Pika 1.0 — Pika Labs · 2023-11 · blog · [video] — https://pika.art/blog
• Adversarial Diffusion Distillation (ADD) — Stability AI · 2023-11 · paper · [method] — https://arxiv.org/abs/2311.17042
• Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets — Stability AI · 2023-11 · paper · [video] — https://arxiv.org/abs/2311.15127
• GenTron: Diffusion Transformers for Image and Video Generation — Meta / HKU · 2023-12 · paper · [method] — https://arxiv.org/abs/2312.04557
• Imagen 2 — Google DeepMind · 2023-12 · blog · [t2i] — https://deepmind.google/technologies/imagen-2/
• Marigold: Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation — ETH Zürich (Photogrammetry and Remote Sensing) · 2023-12 · paper · [method] — https://arxiv.org/abs/2312.02145
• V6.1 — Midjourney · 2023-12 · blog · [t2i] — https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version
• Playground v2 (1024px Aesthetic) — Playground (Playground AI) · 2023-12 · model-card · [t2i] — https://huggingface.co/playgroundai/playground-v2-1024px-aesthetic
• VideoPoet: A Large Language Model for Zero-Shot Video Generation — Google · 2023-12 · paper · [video] — https://arxiv.org/abs/2312.14125
• W.A.L.T: Photorealistic Video Generation with Diffusion Models — Google Research / Stanford University · 2023-12 · paper · [video] — https://arxiv.org/abs/2312.06662
• X-Adapter: Adding Universal Compatibility of Plugins for Upgraded Diffusion Model — Show Lab (NUS) / Tencent AI Lab / Fudan · 2023-12 · paper · [method] — https://arxiv.org/abs/2312.02238

2024（66 条）

• Lumiere: A Space-Time Diffusion Model for Video Generation — Google Research / Weizmann Institute / Tel-Aviv University / Technion · 2024-01 · paper · [video] — https://arxiv.org/abs/2401.12945
• PixArt-δ: Fast and Controllable Image Generation with Latent Consistency Models — Huawei Noah’s Ark Lab · 2024-01 · tech-report · [t2i] — https://arxiv.org/abs/2401.05252
• SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers — New York University (Saining Xie group) · 2024-01 · paper · [method] — https://arxiv.org/abs/2401.08740
• Genie: Generative Interactive Environments — Google DeepMind · 2024-02 · paper · [video] — https://arxiv.org/abs/2402.15391
• Sora: Video generation models as world simulators — OpenAI · 2024-02 · tech-report · [video] — https://openai.com/index/video-generation-models-as-world-simulators/
• Stable Audio: Fast Timing-Conditioned Latent Audio Diffusion — Stability AI (Harmonai) · 2024-02 · paper · [audio] — https://arxiv.org/abs/2402.04825
• Stable Cascade (Würstchen v3) — Stability AI · 2024-02 · model-card · [t2i] — https://stability.ai/news/introducing-stable-cascade
• CogView3: Finer and Faster Text-to-Image Generation via Relay Diffusion — Zhipu AI / Tsinghua University · 2024-03 · paper · [t2i] — https://arxiv.org/abs/2403.05121
• ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment — Tencent · 2024-03 · paper · [method] — https://arxiv.org/abs/2403.05135
• PixArt-Σ: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation — Huawei Noah’s Ark Lab / HKU / DLUT / HKUST · 2024-03 · paper · [t2i] — https://arxiv.org/abs/2403.04692
• MMDiT) — Stability AI · 2024-03 · paper · [t2i] — https://stability.ai/news/stable-diffusion-3-research-paper
• SEED-X: Multimodal Models with Unified Multi-granularity Comprehension and Generation — Tencent ARC Lab / Tencent AI Lab (AILab-CVC) · 2024-04 · paper · [unified] — https://arxiv.org/abs/2404.14396
• Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (VAR) — ByteDance / Peking University · 2024-04 · paper · [method] — https://arxiv.org/abs/2404.02905
• Chameleon: Mixed-Modal Early-Fusion Foundation Models — Meta FAIR · 2024-05 · paper · [unified] — https://arxiv.org/abs/2405.09818
• gpt-image-1) — OpenAI · 2024-05 · blog · [omni] — https://openai.com/index/hello-gpt-4o/
• Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding — Tencent (Hunyuan) · 2024-05 · tech-report · [t2i] — https://arxiv.org/abs/2405.08748
• Lumina-T2X: Transforming Text into Any Modality, Resolution, and Duration via Flow-based Large Diffusion Transformers — Shanghai AI Laboratory & CUHK · 2024-05 · tech-report · [unified] — https://arxiv.org/abs/2405.05945
• Diffusion Loss) — MIT CSAIL / Google DeepMind / Tsinghua (Kaiming He group; Tianhong Li) · 2024-06 · paper · [method] — https://arxiv.org/abs/2406.11838
• Kling (可灵) 视频生成大模型 — Kuaishou (快手) · 2024-06 · blog · [video] — https://ir.kuaishou.com/news-releases/news-release-details/kuaishou-unveils-proprietary-video-generation-model-kling/
• LlamaGen: Autoregressive Model Beats Diffusion — Llama for Scalable Image Generation — University of Hong Kong / ByteDance · 2024-06 · paper · [method] — https://arxiv.org/abs/2406.06525
• Lumina-Next: Making Lumina-T2X Stronger and Faster with Next-DiT — Shanghai AI Laboratory · 2024-06 · paper · [t2i] — https://arxiv.org/abs/2406.18583
• Runway Gen-3 Alpha — Runway · 2024-06 · blog · [video] — https://runwayml.com/research/introducing-gen-3-alpha
• Seed-TTS: A Family of High-Quality Versatile Speech Generation Models — ByteDance · 2024-06 · tech-report · [audio] — https://arxiv.org/abs/2406.02430
• Kolors（可图）: Effective Training of Diffusion Model for Photorealistic Text-to-Image Synthesis — Kuaishou（快手 Kolors 团队） · 2024-07 · tech-report · [t2i] — https://github.com/Kwai-Kolors/Kolors
• Qwen2-Audio — Alibaba (Qwen Team) · 2024-07 · tech-report · [audio] — https://qwenlm.github.io/blog/qwen2-audio/
• UltraEdit: Instruction-based Fine-Grained Image Editing at Scale — PKU / BIGAI / THU · 2024-07 · paper · [edit] — https://arxiv.org/abs/2407.05282
• CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer — Zhipu AI / Tsinghua University · 2024-08 · paper · [video] — https://arxiv.org/abs/2408.06072
• schnell) — Black Forest Labs · 2024-08 · blog · [t2i] — https://bfl.ai/blog/24-08-01-bfl
• Ideogram 2.0 — Ideogram · 2024-08 · blog · [t2i] — https://about.ideogram.ai/2.0
• Imagen 3 — Google DeepMind · 2024-08 · tech-report · [t2i] — https://arxiv.org/abs/2408.07009
• Show-o: One Single Transformer to Unify Multimodal Understanding and Generation — Show Lab (NUS) / ByteDance · 2024-08 · paper · [unified] — https://arxiv.org/abs/2408.12528
• Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model — Meta / Waymo / USC · 2024-08 · paper · [unified] — https://arxiv.org/abs/2408.11039
• Emu3: Next-Token Prediction is All You Need — BAAI（北京智源人工智能研究院） · 2024-09 · tech-report · [unified] — https://arxiv.org/abs/2409.18869
• OmniGen: Unified Image Generation — BAAI（北京智源人工智能研究院） · 2024-09 · paper · [unified] — https://arxiv.org/abs/2409.11340
• Playground v3: Improving Text-to-Image Alignment with Deep-Fusion Large Language Models — Playground · 2024-09 · tech-report · [t2i] — https://arxiv.org/abs/2409.10695
• VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation — NVIDIA / MIT / Tsinghua / UC Berkeley / UC San Diego · 2024-09 · paper · [unified] — https://arxiv.org/abs/2409.04429
• Allegro: Open the Black Box of Commercial-Level Video Generation Model — Rhymes AI · 2024-10 · tech-report · [video] — https://arxiv.org/abs/2410.15458
• Baichuan-Omni — Baichuan Inc.（百川智能，联合西湖大学 / 浙江大学） · 2024-10 · tech-report · [omni] — https://arxiv.org/abs/2410.08565
• FLUX1.1 (pro) — Black Forest Labs · 2024-10 · blog · [t2i] — https://bfl.ai/blog/24-10-02-flux
• HART: Efficient Visual Generation with Hybrid Autoregressive Transformer — MIT / NVIDIA / Tsinghua · 2024-10 · paper · [method] — https://arxiv.org/abs/2410.10812
• In-Context LoRA for Diffusion Transformers (IC-LoRA) — Alibaba Tongyi Lab (ali-vilab) · 2024-10 · tech-report · [method] — https://arxiv.org/abs/2410.23775
• Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation — DeepSeek-AI · 2024-10 · paper · [unified] — https://arxiv.org/abs/2410.13848
• Mochi 1 (preview) — Genmo · 2024-10 · blog · [video] — https://www.genmo.ai/blog/mochi-1-a-new-sota-in-open-text-to-video
• Movie Gen: A Cast of Media Foundation Models — Meta · 2024-10 · tech-report · [video] — https://ai.meta.com/research/movie-gen/
• Pyramid Flow:金字塔式流匹配的高效视频生成 — PKU / Kuaishou / BUPT · 2024-10 · paper · [video] — https://arxiv.org/abs/2410.05954
• REPA: Representation Alignment for Generation — Training Diffusion Transformers Is Easier Than You Think — KAIST / Korea University / Scaled Foundations / NYU · 2024-10 · paper · [method] — https://arxiv.org/abs/2410.06940
• Recraft V3 (red_panda) — Recraft · 2024-10 · blog · [t2i] — https://www.recraft.ai/blog/recraft-introduces-a-revolutionary-ai-model-that-thinks-in-design-language
• SD3.5 Large Turbo 与 (Latent) Adversarial Diffusion Distillation — Stability AI · 2024-10 · paper · [method] — https://stability.ai/news/introducing-stable-diffusion-3-5
• Sana: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformers — NVIDIA / MIT · 2024-10 · tech-report · [t2i] — https://arxiv.org/abs/2410.10629
• Redux) — Black Forest Labs · 2024-11 · blog · [edit] — https://blackforestlabs.ai/flux-1-tools/
• Hunyuan3D 1.0: A Unified Framework for Text-to-3D and Image-to-3D Generation — Tencent · 2024-11 · tech-report · [3d] — https://arxiv.org/abs/2411.02293
• JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation — DeepSeek-AI / PKU / HKU / THU · 2024-11 · paper · [unified] — https://arxiv.org/abs/2411.07975
• LTX-Video: Realtime Video Latent Diffusion — Lightricks · 2024-11 · paper · [video] — https://arxiv.org/abs/2501.00103
• OminiControl: Minimal and Universal Control for Diffusion Transformer — National University of Singapore (xML Lab, Xinchao Wang) · 2024-11 · paper · [edit] — https://arxiv.org/abs/2411.15098
• SeedEdit: Align Image Re-Generation to Image Editing — ByteDance · 2024-11 · tech-report · [edit] — https://arxiv.org/abs/2411.06686
• Aurora (Grok Image Generation) — xAI · 2024-12 · blog · [t2i] — https://x.ai/news/grok-image-generation-release
• GLM-4-Voice: Towards Intelligent and Human-Like End-to-End Spoken Chatbot — Zhipu AI / Tsinghua University · 2024-12 · tech-report · [audio] — https://arxiv.org/abs/2412.02612
• Gemini 2.0 Flash 原生图像生成（Native Image Output） — Google DeepMind · 2024-12 · blog · [omni] — https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/
• HunyuanVideo: A Systematic Framework For Large Video Generative Models — Tencent · 2024-12 · tech-report · [video] — https://arxiv.org/abs/2412.03603
• Infinity: Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis — ByteDance · 2024-12 · paper · [method] — https://arxiv.org/abs/2412.04431
• Liquid: Language Models are Scalable and Unified Multi-modal Generators — HUST / ByteDance / HKU · 2024-12 · paper · [unified] — https://arxiv.org/abs/2412.04332
• MetaMorph: Multimodal Understanding and Generation via Instruction Tuning — Meta (FAIR) / New York University · 2024-12 · paper · [unified] — https://arxiv.org/abs/2412.14164
• Open-Sora（2024 开源复现 Sora） — PKU-YuanGroup（北大-兔展）/ HPC-AI Tech（潞晨） · 2024-12 · tech-report · [video] — https://arxiv.org/abs/2412.00131
• Sora（Sora Turbo）公开发布 — OpenAI · 2024-12 · system-card · [video] — https://openai.com/index/sora-is-here/
• TRELLIS: Structured 3D Latents for Scalable and Versatile 3D Generation — Microsoft Research / Tsinghua / USTC · 2024-12 · paper · [3d] — https://arxiv.org/abs/2412.01506
• Veo 2 — Google DeepMind · 2024-12 · blog · [video] — https://blog.google/innovation-and-ai/models-and-research/google-labs/video-image-generation-update-december-2024/

2025（66 条）

• ACE++: Instruction-Based Image Creation and Editing via Context-Aware Content Filling — Alibaba Tongyi Lab (ali-vilab) · 2025-01 · tech-report · [edit] — https://arxiv.org/abs/2501.02487
• Cosmos World Foundation Model Platform for Physical AI (Cosmos-Predict1) — NVIDIA · 2025-01 · tech-report · [video] — https://arxiv.org/abs/2501.03575
• Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling — DeepSeek-AI · 2025-01 · tech-report · [unified] — https://arxiv.org/abs/2501.17811
• SANA 1.5: Efficient Scaling of Training-Time and Inference-Time Compute in Linear Diffusion Transformer — NVIDIA / MIT / 清华 / Playground / 北大 / HKU · 2025-01 · tech-report · [t2i] — https://arxiv.org/abs/2501.18427
• Fractal Generative Models (FractalGen) — MIT CSAIL / Google DeepMind（何恺明组） · 2025-02 · paper · [method] — https://arxiv.org/abs/2502.17437
• Goku: Flow Based Video Generative Foundation Models — 字节跳动 / 港大 (HKU) · 2025-02 · tech-report · [video] — https://arxiv.org/abs/2502.04896
• 2.2 — Pika Labs · 2025-02 · blog · [video] — https://pika.art/
• Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction — 阶跃星辰 StepFun · 2025-02 · tech-report · [audio] — https://arxiv.org/abs/2502.11946
• CogView4-6B — 智谱 AI / 清华大学 KEG · 2025-03 · model-card · [t2i] — https://github.com/THUDM/CogView4
• 4o 原生图像生成（GPT-4o native image generation） — OpenAI · 2025-03 · system-card · [omni] — https://openai.com/index/introducing-4o-image-generation/
• Gemini 2.0 Flash 原生图像生成（公开实验版 gemini-2.0-flash-exp，2025-03） — Google DeepMind · 2025-03 · blog · [omni] — https://developers.googleblog.com/en/experiment-with-gemini-20-flash-native-image-generation/
• Ideogram 3.0 — Ideogram · 2025-03 · blog · [t2i] — https://about.ideogram.ai/3.0
• Lumina-Image 2.0: A Unified and Efficient Image Generative Framework — 上海AI实验室 / Alpha-VLLM · 2025-03 · tech-report · [t2i] — https://arxiv.org/abs/2503.21758
• Qwen2.5-Omni Technical Report — 阿里巴巴 Qwen 团队 · 2025-03 · tech-report · [omni] — https://arxiv.org/abs/2503.20215
• Reve Image 1.0 (Halfmoon) — Reve AI, Inc. · 2025-03 · blog · [t2i] — https://reve.com/
• VACE: All-in-One Video Creation and Editing — Alibaba Tongyi Lab (Wan Team) · 2025-03 · paper · [video] — https://arxiv.org/abs/2503.07598
• Wan: Open and Advanced Large-Scale Video Generative Models (Wan 2.1) — 阿里巴巴 Wan 团队 (Alibaba Wan Team) · 2025-03 · tech-report · [video] — https://arxiv.org/abs/2503.20314
• 4 Ultra — Adobe · 2025-04 · blog · [t2i] — https://blog.adobe.com/en/publish/2025/04/24/adobe-firefly-next-evolution-creative-ai-is-here
• DreamO: A Unified Framework for Image Customization — ByteDance · 2025-04 · paper · [edit] — https://arxiv.org/abs/2504.16915
• FramePack: Frame Context Packing and Drift Prevention in Next-Frame-Prediction Video Diffusion Models — Stanford University / MIT · 2025-04 · paper · [video] — https://arxiv.org/abs/2504.12626
• ICEdit：用 In-Context 生成范式做指令图像编辑（Enabling Instructional Image Editing with In-Context Generation in Large-Scale DiT） — 浙江大学 ReLER/CCAI · 哈佛 DBMI/HMS · 2025-04 · paper · [edit] — https://arxiv.org/abs/2504.20690
• Lumina-mGPT 2.0: Stand-Alone AutoRegressive Image Modeling — Shanghai AI Laboratory / Alpha-VLLM · 2025-04 · tech-report · [unified] — https://arxiv.org/abs/2507.17801
• MetaQueries: Transfer between Modalities with MetaQueries — Meta / 纽约大学 (NYU) · 2025-04 · paper · [unified] — https://arxiv.org/abs/2504.06256
• Midjourney V7 — Midjourney · 2025-04 · blog · [t2i] — https://www.midjourney.com/updates/v7-alpha
• Seedream 3.0 Technical Report — 字节跳动 Seed · 2025-04 · tech-report · [t2i] — https://seed.bytedance.com/en/seedream3_0
• SkyReels-V2: Infinite-Length Film Generative Model — 昆仑万维 Skywork AI · 2025-04 · tech-report · [video] — https://arxiv.org/abs/2504.13074
• Step1X-Edit: A Practical Framework for General Image Editing — 阶跃星辰 StepFun · 2025-04 · tech-report · [edit] — https://arxiv.org/abs/2504.17761
• UNO: Less-to-More Generalization — Unlocking More Controllability by In-Context Generation — ByteDance Intelligent Creation · 2025-04 · paper · [edit] — https://arxiv.org/abs/2504.02160
• 2.5 Turbo — 快手 Kuaishou（可灵 Kling AI） · 2025-04 · blog · [video] — https://www.nasdaq.com/press-release/kling-ai-advances-20-era-empowering-everyone-tell-great-stories-ai-2025-04-15
• BAGEL: Emerging Properties in Unified Multimodal Pretraining — ByteDance Seed · 2025-05 · paper · [unified] — https://arxiv.org/abs/2505.14683
• BLIP3-o: A Family of Fully Open Unified Multimodal Models — Salesforce Research · 2025-05 · paper · [unified] — https://arxiv.org/abs/2505.09568
• FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space — Black Forest Labs · 2025-05 · tech-report · [edit] — https://bfl.ai/announcements/flux-1-kontext
• HiDream-I1: A High-Efficient Image Generative Foundation Model with Sparse Diffusion Transformer — 智象未来 (HiDream.ai) · 2025-05 · tech-report · [t2i] — https://arxiv.org/abs/2505.22705
• Imagen 4 Fast — Google DeepMind · 2025-05 · blog · [t2i] — https://deepmind.google/models/imagen/
• MAGI-1: Autoregressive Video Generation at Scale — Sand AI · 2025-05 · tech-report · [video] — https://arxiv.org/abs/2505.13211
• MMaDA: Multimodal Large Diffusion Language Models — Princeton University / Peking University / Tsinghua University / ByteDance Seed · 2025-05 · paper · [unified] — https://arxiv.org/abs/2505.15809
• MeanFlow: Mean Flows for One-step Generative Modeling — CMU + MIT (Geng·Deng·Bai·Kolter·He，He 组) · 2025-05 · paper · [method] — https://arxiv.org/abs/2505.13447
• Veo 3 与 Veo 3.1 — Google DeepMind · 2025-05 · blog · [video] — https://blog.google/innovation-and-ai/products/generative-media-models-io-2025/
• Hunyuan3D 2.1: From Images to High-Fidelity 3D Assets with Production-Ready PBR Material — 腾讯混元 (Tencent Hunyuan) · 2025-06 · tech-report · [3d] — https://arxiv.org/abs/2506.15442
• Ming-Lite-Omni: A Unified Multimodal Model for Perception and Generation — 蚂蚁集团 Inclusion AI (Ant Group) · 2025-06 · tech-report · [omni] — https://arxiv.org/abs/2506.09344
• OmniGen2: Towards Instruction-Aligned Multimodal Generation — 北京智源 BAAI · 2025-06 · paper · [unified] — https://arxiv.org/abs/2506.18871
• Ovis-U1: Unified Understanding, Generation and Editing — 阿里巴巴 (Alibaba Group, Ovis Team) · 2025-06 · tech-report · [unified] — https://arxiv.org/abs/2506.23044
• SeedEdit 3.0: Fast and High-Quality Generative Image Editing — ByteDance Seed · 2025-06 · tech-report · [edit] — https://seed.bytedance.com/en/tech/seededit
• Seedance 1.0: Exploring the Boundaries of Video Generation Models — 字节跳动 Seed (ByteDance Seed) · 2025-06 · tech-report · [video] — https://arxiv.org/abs/2506.09113
• Show-o2: Improved Native Unified Multimodal Models — Show Lab (新加坡国立大学) / ByteDance · 2025-06 · paper · [unified] — https://arxiv.org/abs/2506.15564
• UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation — Peking University (PKU-YuanGroup) / Peng Cheng Lab / Rabbitpre AI · 2025-06 · paper · [unified] — https://arxiv.org/abs/2506.03147
• FLUX.1 Krea (dev) — Black Forest Labs / Krea · 2025-07 · blog · [t2i] — https://www.krea.ai/blog/flux-krea-open-source-release
• Wan 2.2 — 阿里巴巴 Wan 团队（Alibaba Wan Team / 通义万相） · 2025-07 · model-card · [video] — https://github.com/Wan-Video/Wan2.2
• X-Omni: Reinforcement Learning Makes Discrete Autoregressive Image Generative Models Great Again — 腾讯混元 (Tencent Hunyuan X) · 2025-07 · paper · [unified] — https://arxiv.org/abs/2507.22058
• Gemini 2.5 Flash Image（Nano Banana） — Google DeepMind · 2025-08 · blog · [edit] — https://blog.google/products-and-platforms/products/gemini/updated-image-editing-model/
• Imagen 4(Ultra) GA 与 GPT Image 1 mini — Google / OpenAI · 2025-08 · blog · [t2i] — https://developers.googleblog.com/en/announcing-imagen-4-fast-and-imagen-4-family-generally-available-in-the-gemini-api/
• NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale — StepFun (阶跃星辰) · 2025-08 · tech-report · [t2i] — https://arxiv.org/abs/2508.10711
• Qwen-Image-Edit-2509 — 阿里巴巴 Qwen 团队 · 2025-08 · model-card · [edit] — https://huggingface.co/Qwen/Qwen-Image-Edit
• Qwen-Image: 阿里巴巴 20B MMDiT 文生图基础模型 — 阿里巴巴 Qwen 团队 · 2025-08 · tech-report · [t2i] — https://arxiv.org/abs/2508.02324
• HunyuanImage 2.1：高效的 2K 高分辨率文生图扩散模型 — 腾讯混元 · 2025-09 · model-card · [t2i] — https://huggingface.co/tencent/HunyuanImage-2.1
• HunyuanImage 3.0 Technical Report — 腾讯混元（Tencent Hunyuan Foundation Model Team） · 2025-09 · tech-report · [unified] — https://arxiv.org/abs/2509.23951
• Seedream 4.0: Toward Next-generation Multimodal Image Generation — 字节跳动 Seed · 2025-09 · tech-report · [unified] — https://seed.bytedance.com/seedream4_0
• Sora 2 — OpenAI · 2025-09 · system-card · [video] — https://openai.com/index/sora-2/
• Diffusion Transformers with Representation Autoencoders (RAE) — 纽约大学 (NYU) / 谢赛宁组 · 2025-10 · paper · [method] — https://arxiv.org/abs/2510.11690
• Emu3.5: Native Multimodal Models are World Learners — 北京智源 BAAI · 2025-10 · tech-report · [omni] — https://arxiv.org/abs/2510.26583
• Lumina-DiMOO: An Omni Diffusion Large Language Model for Multi-Modal Generation and Understanding — Shanghai AI Laboratory / Alpha-VLLM · 2025-10 · tech-report · [unified] — https://arxiv.org/abs/2510.06308
• FLUX.2 — Black Forest Labs · 2025-11 · blog · [t2i] — https://bfl.ai/blog/flux-2
• Gemini 3 Pro Image（Nano Banana Pro） — Google DeepMind · 2025-11 · blog · [edit] — https://blog.google/innovation-and-ai/products/nano-banana-pro/
• HunyuanVideo 1.5 — 腾讯混元 (Tencent Hunyuan Foundation Model Team) · 2025-11 · tech-report · [video] — https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5
• Kandinsky 5.0: A Family of Foundation Models for Image and Video Generation — Sber AI / AIRI (Kandinsky Lab) · 2025-11 · tech-report · [video] — https://arxiv.org/abs/2511.14993
• Z-Image-Turbo：单流扩散 Transformer 的高效 6B 文生图基础模型 — Alibaba Tongyi Lab (Tongyi-MAI) · 2025-11 · tech-report · [t2i] — https://arxiv.org/abs/2511.22699

2026（33 条）

• FLUX.2 (klein) — Black Forest Labs · 2026-01 · model-card · [unified] — https://bfl.ai/blog/flux2-klein-towards-interactive-visual-intelligence
• Luma Ray3.14 — Luma AI · 2026-01 · blog · [video] — https://lumalabs.ai/news/ray3_14
• Skywork UniPic 3.0: Unified Multi-Image Composition via Sequence Modeling — Kunlun Skywork (Skywork Multimodality Team) · 2026-01 · tech-report · [unified] — https://arxiv.org/abs/2601.15664
• Recraft V4 — Recraft · 2026-02 · blog · [t2i] — https://www.recraft.ai/blog/introducing-recraft-v4-design-taste-meets-image-generation
• Seedance 2.0 — ByteDance Seed · 2026-02 · tech-report · [video] — https://seed.bytedance.com/en/blog/official-launch-of-seedance-2-0
• Seedream 5.0 Lite — 字节跳动 Seed · 2026-02 · blog · [unified] — https://seed.bytedance.com/en/blog/deeper-thinking-more-accurate-generation-introducing-seedream-5-0-lite
• Image 3.0 Omni） — 快手（Kuaishou / 可灵 AI） · 2026-02 · blog · [video] — https://kling.ai/release-note/release-notes/whbvu8hsip
• InternVL-U: Democratizing Unified Multimodal Models for Understanding, Reasoning, Generation and Editing — Shanghai AI Laboratory (OpenGVLab) · 2026-03 · tech-report · [unified] — https://arxiv.org/abs/2603.09877
• ChatGPT Images 2.0（含 Thinking mode） — OpenAI · 2026-04 · system-card · [unified] — https://openai.com/index/introducing-chatgpt-images-2-0/
• Midjourney V8.1 — Midjourney · 2026-04 · blog · [t2i] — https://updates.midjourney.com/v8-1-alpha/
• Qwen3.5-Omni Technical Report — Alibaba Qwen · 2026-04 · tech-report · [omni] — https://qwen.ai/blog?id=qwen3.5-omni
• Seed3D 2.0: Advancing High-Fidelity Simulation-Ready 3D Content Generation — ByteDance Seed · 2026-04 · tech-report · [3d] — https://seed.bytedance.com/en/seed3d_2_0
• ERNIE-Image Technical Report — Baidu (ERNIE-Image Team) · 2026-05 · tech-report · [t2i] — https://arxiv.org/abs/2605.25347
• Qwen-Image-2.0 Technical Report — Alibaba Qwen · 2026-05 · tech-report · [unified] — https://arxiv.org/abs/2605.10730
• Qwen-Image-Bench: From Generation to Creation in Text-to-Image Evaluation — Alibaba Qwen · 2026-05 · paper · [method] — https://arxiv.org/abs/2605.28091
• Qwen-Image-VAE-2.0 Technical Report — Alibaba Qwen · 2026-05 · tech-report · [method] — https://arxiv.org/abs/2605.13565
• Ideogram 4.0 — Ideogram · 2026-06 · blog · [t2i] — https://ideogram.ai/blog/ideogram-4.0/
• Krea 2 — Krea · 2026-06 · tech-report · [t2i] — https://www.krea.ai/blog/krea-2-technical-report
• LiveEdit: Towards Real-Time Diffusion-Based Streaming Video Editing — 清华大学 / HKUST · 2026-06 · paper · [edit] — https://arxiv.org/abs/2606.26740
• Nano Banana 2 Lite（Gemini 3.1 Flash Lite Image）与 Gemini Omni Flash 开发者 GA — Google DeepMind · 2026-06 · blog · [unified] — https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-omni-flash-nano-banana-2-lite/
• OmniGen-AR: AutoRegressive Any-to-Image Generation — Fudan University + ByteDance Seed + HKU · 2026-06 · paper · [unified] — https://arxiv.org/abs/2606.09156
• Qwen-Image-2.0-RL Technical Report — Alibaba Qwen · 2026-06 · tech-report · [method] — https://arxiv.org/abs/2606.27608
• Qwen-Image-Flash: Beyond Objective Design — Alibaba Qwen · 2026-06 · tech-report · [t2i] — https://arxiv.org/abs/2606.03746
• Reve 2.0 — Reve AI · 2026-06 · blog · [unified] — https://blog.reve.com/posts/announcing-reve-2.0/
• Muse Image（含 Muse Video 预览） — Meta (Meta Superintelligence Labs) · 2026-07 · blog · [t2i] — https://about.fb.com/news/2026/07/introducing-muse-image-meta-ai/
• Nemotron-Labs-Audex-30B-A3B (Audex): Unified Audio Intelligence Without Regressing on Text Intelligence — NVIDIA · 2026-07 · paper · [omni] — https://arxiv.org/abs/2607.05196
• PixWorld: Unifying 3D Scene Generation and Reconstruction in Pixel Space — 独立研究者 (含 Changhu Wang) · 2026-07 · paper · [3d] — https://arxiv.org/abs/2607.05373
• Qwen-Music Technical Report — Alibaba Qwen · 2026-07 · tech-report · [audio] — https://arxiv.org/abs/2607.11699
• Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image Generation — 字节跳动 Seed · 2026-07 · paper · [method] — https://arxiv.org/abs/2607.11886
• Seedream 5.0 Pro — 字节跳动 Seed · 2026-07 · blog · [unified] — https://seed.bytedance.com/en/blog/beyond-generation-it-understands-design-introducing-seedream-5-0-pro
• Vidu S1: A Real-Time Interactive Video Generation Model — 生数科技 Shengshu · 2026-07 · paper · [video] — https://arxiv.org/abs/2607.03118
• Vision as Unified Multimodal Generation — SenseTime 商汤 (SenseNova) · 2026-07 · paper · [unified] — https://arxiv.org/abs/2607.06560
• Wan-Streamer v0.2: Higher Resolution, Same Latency — 阿里巴巴 Wan / 通义万相 (DAMO) · 2026-07 · tech-report · [video] — https://arxiv.org/abs/2607.04443

未知（11 条）

• Infra：训练规模·并行·tokenizer 工程·推理加速 — - · ? · source · [-] —
• 编辑族横向对比：CogView · Qwen-Image · Hunyuan · Seedream · Kolors · ERNIE 及开源诸侯（2021–2026） — - · ? · source · [-] —
• Qwen-Edit — - · ? · source · [-] —
• 模型族横向对比：Stable Diffusion → SDXL → SD3 → FLUX 谱系 — - · ? · source · [-] —
• 模型架构演进：从 U-Net 扩散到统一 omni 骨干（2020–2026） — - · ? · source · [-] —
• 横向综述 · 数据：规模 · 配比 · 重标注 · 过滤 — - · ? · source · [-] —
• Omni 模型族横向对比：Chameleon · Emu · Janus · BAGEL · OmniGen · Show-o · VAR 谱系 — - · ? · source · [-] —
• 统一理解生成 & any-to-any 全模态专题 — - · ? · source · [-] —
• 视频生成族横向对比（Sora · Veo · Wan · Movie Gen · HunyuanVideo · Kling · CogVideoX 及其谱系） — - · ? · source · [-] —
• 训练方法：目标函数·多阶段·偏好对齐·蒸馏 — - · ? · source · [-] —
• 评测 benchmark 演进与横向数字 — - · ? · source · [-] —