We are happy to launch Anigen V2 today, which fully utilizes the capabilities of the previous model but also adds major developments that allow more anime-accurate characters and styling. It has been in development even before the site was released, but the long time has been worth it.
More Accurate Characters
Anigen V1 can generate accurate characters as well but tends to give messy results when the prompt concatenates too many details within it. Contrary to that, images generated with Anigen V2 become more clear and refined as you enter more details.
This way users who want to create more complex scenarios can create detail-rich artwork, and users who prefer to prompt simply can take advantage of our quality tags option and generate more vivid artwork.
Better Hands
AI art generations can always mess up the hands, especially once your prompt gets complex, but AniGen V2 can generate hands without extra or missing fingers. You don’t need any extra prompting or keywords for it!
Generate More Than One Character
AniGen V2 will produce more consistent results when generating two or more characters in a scene. It effectively separates the prompts based on its training data which allows it to generate multiple characters while maintaining their character attributes.
Subtle Styling
The new AniGen V2 styles are more subtle compared to AniGen V1. But the style the images are generated remains the same.
Official - Similar to the official style of the anime. Better when generating existing anime characters.
AnimeShot - Anime Screenshot style with soft light and lines.
HardMix - More warm colors and hard shadows.
LightMix - More light colors and lighter shadows.
Sketch - Harder pencil strokes.
There will be more changes to these styles in future updates, but if you want a specific style, color, or stroke to be more visible you have the option to switch off all styling and prompt it, or to combine and mix styles according to your preference.
Prompting Changes
Prompting still remains the same in AniGen V2, but now more prompting will result in more details. You also have the ability to apply specific styling (monochrome, manga style, painting) and the model will understand these changes better.
It will also respond well to more background keywords, and make the backgrounds more accurate compared to the limitations of AniGen V1.
AniGen V2 also wouldn’t need a lot of negative prompting to reduce bad-quality images since we have already filtered our datasets and removed a majority of low quality content.
Please expect a prompting guide very soon, which we plan to bundle with our official documentation.
Currently, there will be some slow durations for some users between model switching, but we plan to rectify this very soon! Thank you for your support to AniGen AI and we hope you enjoy the new model! Please don’t hesitate to give your feedback or open any issues via the issues section.
Happy prompting!