Animagine XL3.1, a brand newOpen SourceAnime theme text toimagery model, has been officially released. The release features a series of upgrades and optimizations to the original, giving it a deeper understanding of a wide range of anime works and styles, covering a variety of art styles from ancient to modern.
Animagine XL3.1 extends its understanding of anime works by integrating new datasets. Whether it's a classic work ofup to dateReleased anime are accurately captured and understood by the model. This feature allows Animagine XL3.1 to better reflect the styles and characteristics of various anime when generating images.
Model Download: https://huggingface.co/cagliostrolab/animagine-xl-3.1
Animagine XL3.1 also solves the overexposure problem and improves the quality of the generated images. By optimizing the algorithm and the training process, the model is able to better control the brightness and color of the image when generating the image, thus avoiding the phenomenon of overexposure.
Animagine XL3.1 also adds a new Aesthetics tab and updates the Quality and Year tabs. This allows users to specify specific aesthetic and period styles when generating images, resulting in images that are more responsive to the user's needs.
In order to improve the accuracy of the generated results, Animagine XL3.1 employs label sorting. In this way, the model can generate images more accurately based on the input labels, thus improving the accuracy of the generated results.
On a technical level, Animagine XL3.1 used 2x A10080GB GPUs for about 350 hours of training. ThisadvancedOther hardware devices enable the model to better learn and understand the various details and color representations of anime during the training process.
It is worth mentioning that the Animagine XL3.1 model pre-training uses a dataset containing 870,000 ordered and labeled images. These images cover a wide range of anime characters, styles, and themes, providing a deep knowledge base for the model.