Recently, Bilibili in China launched an open-source animation video generation model called “Index-AniSora,” which can generate anime-style video shots with just one click. Bilibili stated that Index-AniSora can fill the gaps in current video generation models when it comes to handling anime videos, and it can better understand the trajectories of animation movements.
Bilibili launches open-source animation video generation model Index-AniSora, strengthening AI anime video generation capabilities.
Bilibili recently released an open-source animation video generation model called “Index-AniSora.” Index-AniSora can generate video shots in various anime styles, including series episodes, original animations, manga adaptations, VTuber content, anime PVs, and more. According to the model overview provided by Index-AniSora, it collects a vast amount of animation resources such as anime series, films, self-made animations, MMD, and virtual idol videos, then cuts these contents into shots and filters out low-quality segments. The filtered clips then become the training dataset for Index-AniSora.
Index-AniSora states that while advanced video generation models like Sora and Kling have succeeded in generating realistic videos, anime videos have a unique style that sometimes goes against the laws of physics in real life. As a result, Sora and Kling are less effective when it comes to anime image generation. In contrast, the Index-AniSora video generation model proposes AniSora, a comprehensive system specifically designed for anime video generation, which handles anime image generation very well, such as in generation across different anime art styles:
- The figure in the painting sits in a car moving forward, waving toward the rear, their hair swaying from side to side in the wind.
- In the picture, two figures in red wedding dresses hold a red string between them, walking off into the distance.
.gif)
- The yellow-haired figure reached out to touch the head of the kneeling figure, whose body heaved as it gasped for breath.
- In the scene, a person sprints forward at high speed, and because the speed is so great, their movements appear somewhat blurred.
- In the shot, the character raises their arm, with airflow clearly flowing along the surface of the arm.
Everyone can GitHub and Hugging Face After seeing the detailed introduction to Index-AniSora, you can even go to the Komiko video generation website to experience AniSora’s video generation effects. However, using AniSora on Komiko requires payment, so you can decide for yourselves. If you’re interested, click the link below to visit Komiko for more details.
Go to Komiko to experience Index-AniSora video generation effects.
Source: KOCPC Chinese