ByteDance releases SDXL-Lightning, an open model for text and graphics, which increases the generation speed tenfold

According to Interface News, sources close to the matter have learned that theByteDancePublishedVincennes diagram open model SDXL-Lightning. is said to be capable of generating high-quality and high-resolution images in a very short period of time, making it one of the fastest Venn diagram models available.

ByteDance releases SDXL-Lightning, an open model for text and graphics, which increases the generation speed tenfold

Venn diagram is a technique that utilizes artificial intelligence to generate images based on textual descriptions. Currently, the mainstream models in the field of text-generated maps use the generation technique of diffusion process, i.e., the noise is gradually transformed into an image through multiple iterations. Although this technique is capable of generating realistic images, it also has the disadvantages of high consumption of computational resources and slow generation speed. It takes about 5 seconds to generate a high-quality image.

ByteHop's SDXL-Lightning model, on the other hand, employs a technique of progressive anti-distillation to achieve unprecedented generation speeds, which is capable of generating images of extremely high quality and resolution in 2 or 4 steps.Accelerates generation by a factor of ten, making it the fastest Venn diagram model at 1024 resolution, and reduces computational cost to one-tenth.

According to the ByteDance Intelligent Creation team, the model is based on an improved version of the open-source Venn diagram model SDXL, compatible with other tools and plug-ins in the open modeling community, SDXL-Lightning can be seamlessly integrated as a speed-up plug-in for SDXL models of diverse styles such as cartoon, anime, and so on, and it supports the current popular control plug-in ControlNet, the generation software The SDXL-Lightning also supports the popular control plug-in ControlNet and generation software ComfyUI, making it easy for developers, researchers, and creative practitioners to use these tools in conjunction with each other to help the entire industry innovate and collaborate.

The model has been published in the AI open source community Hugging Face on the public, among the top model trends, and also became a popular model on Hugging Face Spaces.

statement:The content is collected from various media platforms such as public websites. If the included content infringes on your rights, please contact us by email and we will deal with it as soon as possible.
Information

Microsoft Copilot chatbot opens "notebook" feature with up to 18,000 characters

2024-2-24 8:17:23

Information

Using AI to restore ancient poetry, China's first AI animation video "Poetry of the Thousand Years" will be broadcast on CCTV on February 26

2024-2-24 8:19:38

Search