Stability AI has recently launched Stable Diffusion 3.5, an advanced text-to-image AI model that offers multiple model variants for commercial and non-commercial use. These models are designed to run on consumer-grade hardware and are available under the flexible Stability AI Community License. This allows developers to customize and integrate the models without worrying about restrictive licensing, making them suitable for a wide range of applications.
The Stable Diffusion 3.5 Large and Stable Diffusion 3.5 Large Turbo models can be downloaded from Hugging Face and the inference can be accessed on GitHub. This latest release offers a range of models catering to different users, including researchers, startups, and enterprises. The Stable Diffusion 3.5 Large model, with 8 billion parameters, delivers superior image quality and prompt adherence, making it ideal for professional use at a 1-megapixel resolution. The Large Turbo version is a faster alternative, generating high-quality images in just 4 steps. The company claims that the model is also optimized for efficient performance on standard consumer hardware, particularly in the Medium and Large Turbo versions. Additionally, it generates inclusive and diverse images, accurately representing various skin tones and features without needing extensive prompts.
The models are trained on a subset of the LAION-5b dataset, created by the DeepFloyd team, and further filtered using the dataset’s NSFW filter to remove adult content. The model is available at no cost for non-commercial purposes, including academic research. Startups, small to medium businesses, and creators can use the model commercially for free, provided their annual revenue is under $1M. Users maintain full ownership of the generated content, with no restrictive licensing.
In contrast, Google has recently announced that it is pausing its Gemini artificial intelligence image generation feature due to inaccuracies in historical pictures. The Gemini-generated images went viral on social media, leading to widespread criticism and anger. Some users have accused Google of sacrificing truth and accuracy in favor of being socially aware. In response, Google stated that they have decided to pause the image generation of people while they work on improving the accuracy of its responses.
The company has been facing backlash for the AI tool’s tendency to generate images of historical figures, such as the U.S. Founding Fathers, as people of color, which many have deemed inaccurate. This decision by Google highlights the importance of ensuring accuracy and inclusivity in AI models, and Stability AI’s Stable Diffusion 3.5 aims to address these concerns by generating diverse and accurate images.
In conclusion, Stability AI’s Stable Diffusion 3.5 is a significant advancement in text-to-image AI models, offering multiple variants for different use cases and addressing concerns of accuracy and inclusivity. With its flexible licensing and availability for both commercial and non-commercial use, this latest release is set to make a significant impact in the AI community.
