即梦AI肢体与动态场景环境
#肢体与动态#场景环境
中文翻译
我们回来了……同样的预算,要么做一个 Seedance 2.5 片段,要么做约 77 个 W1 片段
在同等时长视频的给定费率下,这就是差距。@VivixLabs_HQ W1 的运行成本为每秒 $0.003,而一段 8 秒视频大约 8 秒就能渲染完成
在并排对比中,视觉效果也经得起考验
而且这个价格比看起来更重要。W1 是一个实时交互式视频模型,而只有每一秒都足够便宜,视频才能在你与之交互时持续生成。现在它做到了
在复杂运动和密集场景方面仍有差距,全面训练正在进行中。但在这个价格下,我会停止挑选简单的提示词。给它一个所有人都在移动的拥挤场景,看看它会在哪里崩掉
现在可通过 API 实时使用,浏览器中暂时只能生成完成的片段。实时浏览器工具即将推出 👇
原帖
we're so back.. one Seedance 2.5 clip or ~77 W1 clips for the same budget
that's the gap at the supplied rates for equal-length videos. @VivixLabs_HQ W1 runs at $0.003/sec, and an 8s video renders in ~8s
the visuals hold up in the side-by-side comparisons too
and that price matters more than it looks. W1 is a real-time interactive video model, and video that keeps generating while you interact with it only works if every second is cheap. now it is
still gaps in complex motion and dense scenes, and full-scale training is underway. but at this price, i'd stop picking easy prompts. give it a crowded scene with everyone moving and see where it falls apart
real-time via API now, finished clips in the browser for now. real-time browser tools coming soon 👇
