ByteDance launches Seedream 5.5 for pixel-stable multi-round image edits
It is rolling out on Jimeng, Doubao, Doubao Work, and Volcengine Ark, with a Pro version planned.
Stories per day · 30 days
4 RSS
Relateddiffusion7benchmark5text-to-image5comfyui4distillation4huggingface4multimodal4canva3diffusion-models3flux3imagenet3lora3open-weights3canva-craft2gpt-image2leonardo-ai2minimax2moss-image2
89 stories, newest first · one card per story
Stories tagged #image-generation
Patch-wise PCA links v-prediction failure to a bandwidth bottleneck
A Raw-injected Transformer restores effective v-prediction and is competitive on ImageNet 256×256.
Comfy-Org republishes MiniMax-H3-Image weights for ComfyUI
The pack adds int8 and NVFP4 files plus Qwen encoders, under a license that restricts high-revenue hosted use.
dNaViT encodes images as discrete general-purpose visual tokens
Authors say it matches or beats task-specific baselines on 12 understanding, 4 generation, and one editing benchmark.
OpenAI Launches GPT-Image-2.6 Flare and Sunburst
New models add GIFs, SVG export and six-second generation time.
WeLM-Image unified image generation model integrates multimodal understanding
WeLM-Image combines text-to-image generation, instruction-guided editing and multi-reference synthesis on a single GPU with MoT backbone and MoE modules, ranking first on PRISM-Bench and Qwen-Image-Bench.
MiniMax releases open-weight MiniMax-H3-Image for generation and editing
The checkpoint outputs about 16-megapixel images in 6 steps and accepts up to 16 reference images.
LARC advances continuous-space reasoning in multimodal models
LARC reasons over continuous hidden states to fix token bottlenecks, boosting image generation quality and test-time scaling on GenEval.
sDMD adds adaptive Sinkhorn correction to diffusion distillation
With 2–4 steps on FLUX.1-dev, Qwen-Image-20B, and Wan2.1-14B, authors report gains over DMD2 and other baselines.
Recipe adapts pretrained autoregressive image models to any token order
On LlamaGen and RAR, the method beat MaskGIT inpainting FID 6.21 to 10.44 at 13-33% extra compute.
SpecSlot cuts global attention cost in pixel-space transformers
Only the lowest-frequency slot per spatial anchor attends globally; higher-frequency detail stays local.
PhysInsert inserts objects with geometry and scene interaction
The authors built Insert World and PhysInsert-Bench and report better fidelity plus a user study on placement and lighting.
MoI-Bench scores mobile image generation and editing gaps
The suite has 64 subtasks and 1,150 samples, with MoI-Eval scoring instruction following, quality and detail preservation.
Black-box signatures proposed to detect LoRA adaptations of diffusion models
Tests use only the intrinsic dimension of generated images and assume unknown adaptations and model families.
FrameForge Uses Video Model Priors for Unified Image Generation
FrameForge models condition-target relationships with Mixture of Temporal Experts and Chain of Temporal Reasoning for superior subject-driven generation.
NLD-Uni unifies visual understanding and generation in a diffusion LLM
Cross-paradigm distillation and progressive token corruption are used instead of conventional masked diffusion.
Wiener-informed decoding speeds pixel-space flow model training
On ImageNet 256, PixelDiT-XL with WiD reports FID 1.53 after 500 epochs, matching prior quality with 37.5% fewer epochs.
Canva releases Craft, its first in-house image model
Craft plans composition and text, renders up to 2K across more than 30 aspect ratios, and is in Dream Lab and a public API.
Canva launches Craft image model that plans designs first
Craft is available globally in Dream Lab, outputs up to 2K across more than 30 aspect ratios, and ranks fourth on Design Arena.
1
Canva releases Canva Craft, its first image generation model
Canva Craft plans a layout first, then renders up to 2K in 30 aspect ratios, and is open to all users.
1
T-DIG lets users steer image diversity with natural-language prompts
It keeps the image generator frozen and adds user-directed diversity to open and closed models without weight access.
MindView benchmark tests generative multi-view spatial intelligence
Fine-tuning BAGEL on MindView raises its overall score by 24.3 points and improves all seven tasks.
Kromoss fine-tune of MOSS-Image-Base trends on Hugging Face
The 1024 px checkpoint ships two 8-step turbo LoRAs and is presented as a drop-in for MOSS-Image-SFT.
6
RoAR-VAE improves latent diffusability via robust prediction
It beats the strongest baseline tokenizer by 2.19 points and FLUX.1-VAE small-text accuracy by 16.93 points.