Danbooru Dataset Filter: Fast local metadata-based search across 10M+ images for LoRA/Checkpoint training
https://redd.it/1sl8cqi
@rStableDiffusion
And most of the nodes are bright red
https://redd.it/1slgfq2
@rStableDiffusion
Ostris AI Toolkit has day zero support for training LoRAs on top of Baidu's ERNIE-Image
https://redd.it/1slivar
@rStableDiffusion
Media is too big
VIEW IN TELEGRAM
Tencent HY-World 2.0 appears to be dropping on April 15 — open-source multimodal 3D world generation from Tencent Hunyuan

https://redd.it/1sll638
@rStableDiffusion
Nucleus-Image Released

https://huggingface.co/NucleusAI/Nucleus-Image
Nucleus-Image is a text-to-image generation model built on a sparse mixture-of-experts (MoE) diffusion transformer architecture. It scales to 17B total parameters across 64 routed experts per layer while activating only \~2B parameters per forward pass, establishing a new Pareto frontier in quality-versus-efficiency. Nucleus-Image matches or exceeds leading models including Qwen-Image, GPT Image 1, Seedream 3.0, and Imagen4 on GenEval, DPG-Bench, and OneIG-Bench. This is a base model released without any post-training optimization (no DPO, no reinforcement learning, no human preference tuning). All reported results reflect pre-training performance only. We release the full model weights, training code, and dataset, making Nucleus-Image the first fully open-source MoE diffusion model at this quality tier.

https://redd.it/1slpfch
@rStableDiffusion