An early version of wan2.2 high-noise training, used in conjunction with the low-noise version. Experimental and may be unstable.
此LoRA模型仅供社区内部进行技术交流与学习研究使用,严禁任何形式的商业用途。
禁止转售或搬运: 未经许可,严禁将本模型文件进行任何形式的转售、分享或上传至其他平台。
禁止商业用途: 禁止将本模型生成的任何图像、视频、动画、游戏资产等,出售、出租、广告宣传、出版物等商业活动。
法律责任: 一经发现上述侵权行为,将保留追究其法律责任的权利。
商业合作: 如有任何商业使用需求,请务必通过联系我进行授权购买。
感谢您的理解与配合。(适用于wan2.1和wan2.2)
An early version of high-noise training, used in conjunction with the low-noise version. Since high-noise model training is still in the exploratory phase, instability may occur.
wan2.2基于低噪模型训练,使用时需要高低噪模型同时加载,并且高噪权重为2,低噪权重为1.测试环境基于kijai-wanvideo2_2_I2V_A14B_example_WIP修改。
wan2.2 is trained based on a low-noise model. When used, both the high-noise and low-noise models must be loaded simultaneously, with the high-noise weight set to 2 and the low-noise weight set to 1. The test environment is modified from kijai-wanvideo2_2_I2V_A14B_example_WIP.
Prompt word examples
H-gfbz-2.2,女人向前行走,镜头跟随然后慢慢拉远,女人向前行走的同时服装逐渐变成粉白配色的草莓主题蛋糕裙,蓬松的裙摆由三层薄纱叠加,每层边缘缝制奶油蕾丝花边。上衣是纯白方领短袖,胸口点缀立体草莓刺绣与珍珠纽扣,腰间系着红色蝴蝶结缎带。搭配白色蕾丝长袜与红色玛丽珍鞋,头戴草莓发箍,手持迷你草莓蛋糕手拿包,整体甜美如童话里的点心精灵,变身完成后用双手对镜头比心
这是一个万相图生视频的变化lora,可以进行人物变装,形象变化,例如变动物,还支持首尾帧变化
测试环境 kijai工作流,wan-i2v-14B-720P,wan-vace-Start and end frames
提示词示例
1,在街道上,一位穿着白色T恤牛仔短裙的女人,在向前行走的同时服装变成淡粉色汉服,gfbz效果。
2,宁静的中国寺庙院落,一个穿着袈裟的和尚站在一颗巨大的树下面,和尚向前行走的同时逐渐变成一只大熊猫,gfbz效果。
This is the wan2.1-i2v Infinite Variations Lora.. Currently, it has been tested and supports character costume changes, transformations (e.g., into animals), and also changes between the first and last frames.
Testing Environment: Kijai workflow, wan-i2v-14B-720P,wan-vace-Start and end frames
Example Prompts:
A woman in a white T-shirt and denim skirt is walking down a street. As she walks forward, her clothes transform into a light pink Hanfu. gfbz effect.
In the serene courtyard of a Chinese temple, a monk in a kasaya robe stands under a giant tree. As the monk walks forward, he gradually transforms into a giant panda. gfbz effect.
祝大家玩的开心,I wish you all a happy time,
An early version of wan2.2 high-noise training, used in conjunction with the low-noise version. Experimental and may be unstable.
1. The rights to reposted models belong to original creators.
2. Original creators should contact SeaArt.AI staff through official channels to claim their models. We are committed to protecting every creator's rights. Click to Claim
