Opening Twitter and the diffusers github every 15 minutes to check for updates on the flux training code.

Unity presents IPAdapter-Instruct Resolving Ambiguity in Image-based Conditioning using Instruct Prompts discuss: huggingface.co/papers/2408.03… Diffusion models continuously push the boundary of state-of-the-art image generation, but the process is hard to control with any nuance:

IPAdapter-Instruct, a new release led by CiaraRowles Unity for Games Feed the IPAdapter image and instruct the model what to use from it: style, color, composition, pose, face! SD1.5, XL and 3 support 🔥 ✴️ Weights: huggingface.co/CiaraRowles/IP… ▶️ Demo: huggingface.co/spaces/unity/I… 👩‍💻

If the berry stuff is real this will have been the weirdest guerilla marketing campaign in history, I fully expect to see anonymous mcdonalds representatives promising salvation in the form of their next burger but in a really vague way through unconnected accounts within a year.

Unity has just released their IP-Adapter for Stable Diffusion models, including SD 1.5, SDXL, and SD3! This innovative tool combines natural-image conditioning with Instruct prompts, allowing you to seamlessly swap between interpretations for the same conditioning image -

Is AI eating all the energy? How expensive are LLMs and image generators? Energy stuff! Part 1 of my two-part report that was going to be a single article but then it got too long. blog.giovanh.com/blog/2024/08/1…

We've just put out an awesome new paper on text-to-3D using Stable Diffusion, Collaborative Control and Geometry Images! check it out! :D

As a Brit, hearing nick clegg act as meta spokesperson feels weird, like if Hilary Clinton suddenly became a PR manager at Microsoft.

Kinda surprised there's no easy way to use ic relight but toned down so it preserves more of the condition image, i wish there was a paper for this to read.