görselmodel·arxiv-cv·20 saat önce·+1 kaynak
Multimodal Diffusion Transformers (MM-DiTs) jointly process visual and textual representations throughout generation. These models repeatedly update the text tokens through multimodal attention, forming dynamic contextual tokens whose function is not well understood. In this work, we introduce a framework for reading this contextual space through natural-language interrogation. We train a lightweight bottleneck network that maps intermediate contextual tokens into the input space of a frozen Large Language Model (LLM), allowing the LLM to answer questions about the emerging image directly…
görsel·reddit-stablediffusion·2 saat önce
I was tired of opening heavy photo editors just to tweak lighting or color-grade a folder of AI generations. So I built TinyLuma - an instant, lightweight photo editor designed specifically for finishing AI artwork and photo sets. No installation required, no subscriptions, and the whole app is only ~8 MB. What’s inside: - All essential controls: Clean sliders for Light (Exposure, Contrast, Highlights, Shadows, Whites, Blacks), Color (Temp, Tint, Vibrance, Saturation), plus Dehaze and natural Film Grain. - Crisp details & texture: Custom Clarity, Texture, and Sharpen sliders to enhance…
görsel·reddit-stablediffusion·3 saat önce
Hey guys! I’m pretty new to AI image generation, and I’ve been experimenting for about 3 months trying to get the best realistic Pinterest-style images. I’m currently using my character LoRA + Krea 2 on an RTX 5060 8GB with 16GB DDR3 RAM. I started with Koo’s workflow from Discord and tweaked it a bit for my own setup. From what I’ve heard, Chroma is pretty good for experimenting with camera angles, compositions, and generating more random/varied images, which is exactly what I’m trying to achieve. So I tried building a workflow where I: - Generate random Pinterest-style prompts using…