Researchers have developed Flash Diffusion, a fast 4-step version of the SD3 model, capable of generating 1024x1024 images in seconds. The model has received praise for its speed and creativity boost in image generation.
Totally amazed by this Space! You can iterate your image ideas on the fly, a huge boost for creativity! (real time video) Flash Diffusion and Tiny AutoEncoder for Stable Diffusion 3 https://t.co/8qCgMwa2Jq Thanks @radamar https://t.co/Hdma3eaoN1
A few days later, we have Flash Diffusion and Tiny AutoEncoder for Stable Diffusion 3, making inference extremely fast! This is an incredibly quick turnaround from the researchers. P.S. The video is not sped up. demo: https://t.co/jRxl2vYXTl https://t.co/1kiTWSz8wm
Super duper quick, after SD3, a distilled student model called Flash Diffusion has retrained on SD3 to bring you an extremely fast 4 steps version that looks quite incredible! 🔥 Kudos to the team, give this a try folks 👇 https://t.co/6szOcUWFg0
Congrats on the flash release of SD3-Flash 📸🏃 Less than 1 week after the model is out there's already a way to generate images in < 2s 🤯 https://t.co/2UYeoIQ4bg https://t.co/RmKeVzEuGa
⚡Flash Diffusion: FlashSD3⚡ - Flash Diffusion, a diffusion distillation - Accelerate any conditional diffusion model for Few-steps Image Gen - 90.4M LoRA distilled version of SD3 - Can generate 1024x1024 images in 4 steps https://t.co/xwQPZbqzWA