2023年被誉为AI视频的突破性一年。年初时，还没有公开的文本转视频模型，但仅仅12个月后，数十种视频生成产品已经被广泛使用，全球数百万用户通过文本或图像提示创建短片。这一进展无疑令人瞩目，然而，这些产品仍然存在一些局限性。
大多数生成的视频时长仍然只有3到4秒，而且输出的质量也参差不齐。角色一致性等问题也还未得到解决。尽管如此，AI视频的发展潜力仍然巨大。Justine Moore，A16Z的合伙人，提到，我们距离用单一文本提示创造出皮克斯级别的短片还有很长的路要走。
这一领域的发展不仅需要技术的突破，还需要解决一系列的挑战。如何提高视频的输出质量，如何解决角色一致性问题，如何延长视频的时长，这些都是需要解决的问题。
尽管存在挑战，但AI视频的发展前景仍然广阔。有新Newin的数据显示，AI视频领域的投资也在不断增加，这将进一步推动这个领域的发展。
English Translation:
The year 2023 is heralded as a breakthrough year for AI video. At the beginning of the year, there were no publicly available text-to-video models, but just 12 months later, dozens of video generation products have been widely adopted, with millions of users around the world creating short videos through text or image prompts. This progress is undoubtedly impressive, yet these products still have some limitations.
Most generated videos still only last 3 to 4 seconds, and the output quality varies. Issues like character consistency have not yet been resolved. However, the potential of AI video is still vast. Justine Moore, a partner at A16Z, mentioned that we still have a long way to go to create Pixar-level shorts from a single text prompt.
The development of this field not only requires technological breakthroughs but also needs to address a series of challenges. How to improve the output quality of videos, how to solve issues of character consistency, and how to extend the duration of videos are all problems that need to be addressed.
Despite these challenges, the future of AI video is still promising. Data from Newin shows that investment in the AI video field is also increasing, which will further drive its development.
【来源】https://mp.weixin.qq.com/s/_WdPoPfptK8zch7C4dh1iQ
Views: 1