SDXL-Turbo: A faster, sharper open-source local AI image generator

Haebom

Nov 29, 20232y ago

Stability AI has once again unveiled something incredible. After accelerating the open source ecosystem for image generation with Stable Diffusion, they've now introduced SDXL-Turbo, combining the speed of SDXL, which was released earlier this year. They've managed to reduce the image generation time of the original SDXL by over 30% on average. (Depending on your computer specs)

stabilityai/sdxl-turbo · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co

SDXL Turbo

youtu.be

Clipdrop - SDXL Turbo

clipdrop.co

You can try it out directly on Clipdrop above, or download the model from Hugging Face and run it locally. As always, it's open source and free. This is the first case where ADD (Adversarial Diffusion Distillation) technology has been released not just as a paper, but as an actual model and product.

adversarial_diffusion_distillation.pdf7.70MB

It's amazing that they've nailed both speed and quality. Previously, Diffusion methods handled noise step-by-step—100 steps, then SDXL brought it down to 50—but now they've reduced it to just 1 step, which is truly impressive. 'Can this really work?' It's fascinating how they've approached it. I think we're about to see a lot of cost savings and fresh new ideas in image generation. On top of that, they've released a technology that can turn images into videos, so I bet we'll soon see their creations popping up everywhere on TikTok, YouTube, and Meta.

Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets — Stability AI

We present Stable Video Diffusion — a latent video diffusion model for high-resolution, state-of-the-art text-to-video and image-to-video generation.

stability.ai

Stable Audio: Fast Timing-Conditioned Latent Audio Diffusion — Stability AI

Stable Audio represents the cutting-edge audio generation research by Stability AI’s generative audio research lab, Harmonai. We continue to improve our model architectures, datasets, and training procedures to improve output quality, controllability, inference speed, and output length.

stability.ai

Subscribe to 'haebom'

📚 Welcome to Haebom's archives.
---
I post articles related to IT 💻, economy 💰, and humanities 🎭.
If you are curious about my thoughts, perspectives or interests, please subscribe.
haebom@kakao.com