返回顶部
热门问答 更多热门问答
技术文章 更多技术文章

Stable Diffusion 3:我们将在本月 12 日开源

[复制链接]
链载Ai 显示全部楼层 发表于 4 小时前 |阅读模式 打印 上一主题 下一主题


长话短说

SD3Medium(2B的那个),下周三起,将通过 Hugging Face 提供下载



官方致信

通过邮件的方式,发给了各位开发者

(中文是我翻译的)


Hey 大聪明,

Have you heard that the SD3 weights are dropping soon? Our co-CEO Christian Laforte just announced the weights release at Computex Taipei earlier today.

你听说了吗?SD3 权重即将发布!我们的联合 CEO Christian Laforte 今天早些时候在台北国际电脑展上宣布了这一消息。

Stable Diffusion 3 Medium, our most advanced text-to-image is on its way! You will be able to download the weights on Hugging Face from Wednesday 12th June.

我们最先进的文本生成图像模型 Stable Diffusion 3 Medium 即将上线!你可以在 6 月 12 日星期三在 Hugging Face 上下载这些权重。


SD3 Medium is a 2 billion parameter SD3 model, specifically designed to excel in areas where previous models struggled. Here are some of the standout features:

SD3 Medium 是一个拥有 20 亿参数的 SD3 模型,专为解决之前模型的不足而设计。以下是一些亮点功能:


  • Photorealism: Overcomes common artifacts in hands and faces, delivering high-quality images without the need for complex workflows.

    • 逼真度:解决了手部和面部常见的瑕疵问题,生成高质量图像,无需复杂的工作流程。

  • Typography: Achieves robust results in typography, outperforming larger state-of-the-art models.

    • 字体设计:在字体设计方面表现出色,超越了更大规模的先进模型。

  • Performance: Ideal for both consumer systems and enterprise workloads due to its optimized size and efficiency.

    • 性能:由于其优化的体积和效率,适用于消费者系统和企业工作负载。

  • Fine-Tuning: Capable of absorbing nuanced details from small datasets, making it perfect for customization and creativity.

    • 微调:能够从小数据集中吸收细微的细节,非常适合定制和创意。


SD3 Medium weights and code will be available for non-commercial use only. If you would like to discuss a self-hosting license for commercial use of Stable Diffusion 3 please complete the form below and our team will be in touch shortly.

SD3 中等权重和代码仅供非商业用途。如果您想讨论 Stable Diffusion 3 的商业自托管许可证,请填写以下表格,我们的团队会尽快与您联系。


From,

Your friends at Stability AI

来自,

您在 Stability AI 的朋友


其他


先补两个注释:

  • 文中提到的 Computex Taipei,昨天英伟达老黄发表了专场演讲,并发布了一些东西:英伟达发布会:揭开新工业革命序幕

  • SD 的商业许可的申请地址是:https://stability.ai/enterprise


以及,当前开放的版本是 2B 版,希望后面可以有 8B 版的到来。对此,Reddit 上有讨论到:

I really hope we get the 8B but if the artisan channels on their Discord are anything to go by it still needs quite a bit of work, especially anatomy. Hopefully we can at least get a clear answer of whether or not they still intend to release it. This "2B is all you need!!!" crap being pushed by multiple employees smells like complete gaslighting. There are numerous projects, like Pony Diffusion, that intend to train on the biggest SD3 model available.

我真的很想要 8B,关于这个事儿希望有一个明确的答复,无论是否仍打算发布。但“2B 就够了!”的信息却被多位工作人员转发。有许多项目,比如 Pony Diffusion,打算在最大的 SD3 模型上进行训练。


Meta, for example, told the community that Llama 405B will take longer to train but will still be released openly. It's not that hard, it's a single sentence but clears everything up. Just say "8B will get a local release" or "8B is reserved for API only".

可以看一下 Meta,他们告诉社区,Llama 405B 将需要更长时间来训练,但仍将公开发布。这并不难,只是一句话就把一切都澄清了。只要说“8B 将会本地发布”或“8B 仅供 API 使用”。

回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则

链载AI是专业的生成式人工智能教程平台。提供Stable Diffusion、Midjourney AI绘画教程,Suno AI音乐生成指南,以及Runway、Pika等AI视频制作与动画生成实战案例。从提示词编写到参数调整,手把手助您从入门到精通。
  • 官方手机版

  • 微信公众号

  • 商务合作

  • Powered by Discuz! X3.5 | Copyright © 2025-2025. | 链载Ai
  • 桂ICP备2024021734号 | 营业执照 | |广西笔趣文化传媒有限公司|| QQ