即梦AI肢体与动态
#肢体与动态
中文翻译
每一个“真正的好镜头”背后,有时都有几十个“差点就成好镜头”的镜头被剪掉,躺在剪辑室的地板上!
看看这个里面法国先生冻僵的手和脸。
我注意到你得真的提醒Seedance,你想要持续生动、富有表现力的面孔,否则它很容易默认呈现出非常僵硬、机械的感觉。
与一些其他视频模型相比,比如备受喜爱的Grok Imagine,它在微妙情感和微表情方面做得非常好,但另一方面,它永远无法在一个30秒的视频中达到这种连贯的动作或运动水平,更不用说15秒了。希望这个差距能很快缩小!
正在制作一些非常有趣的新东西——很兴奋能分享更多。
原帖
for every "Really Good Take" there's sometimes dozens and dozens of "Almost Really Good" takes left on the cutting room floor!
see mr frenchman's frozen hand and face in this one.
I've noticed you gotta really remind Seedance that you want continuously alive and expressive faces otherwise it loves to default to a very rigid and robotic feel.
Compared to some other video models like the beloved Grok Imagine which do subtle emotions and micro expressions very well, but on the other hand could never get this level of coherent action or motion in a single 30 second video much less 15 seconds. Hopefully soon that gap will close!
Working on some very fun new stuff - excited to share more.
