home.social

#diffusionmodel — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #diffusionmodel, aggregated by home.social.

  1. Steerling-8B, the first interpretable model that can trace any token it generates to its input context, concepts a human can understand, and its training data.

    guidelabs.ai/post/steerling-8b

    #AI #InterpretableAI #DiffusionModel #DiffusionModels

  2. Alibaba’s new Qwen‑Image‑2.0 diffusion model can render Chinese calligraphy with near‑perfect text, edging out most generators except the quirky Nano Banana Pro. It shines in multimodal generation and image editing. See how open‑source fans are reacting. #QwenImage2 #AIImageGeneration #ChineseCalligraphy #DiffusionModel

    🔗 aidailypost.com/news/qwen-imag

  3. This is the next stage: W.A.L.T, a #DiffusionModel for #photorealistic #VideoGeneration 🤖🎬 Developed by Gupta et al. (2023), based on a #transformer that is trained on image and video generation in a shared #LatentSpace.

    🌍 walt-video-diffusion.github.io (including many sample movies)
    📔 arxiv.org/abs/2312.06662

    #WALT