Text-to-video models already exist, but none have worked as well as OpenAI’s Sora
OpenAI , the developers behind ChatGPT, are not sleeping and are working on more and more AI tools. Now they have presented the new text-to-video model Sora
According to the (official website) Sora is designed to create “realistic and imaginative scenes” from a single prompt.
The biggest difference to other text-to-video models clearly lies in the precise representation of content, which has been in constant flux in most AI videos to date.
Sora: Realistic videos in the blink of an eye
What can Sora do? OpenAI’s latest AI achievement, Sora, promises to revolutionize the way artificial intelligence understands the physical world – or at least take it a whole step forward.
With the ability to create scenes with multiple people, different types of movement and detailed objects, Sora sets itself apart from other AI platforms.
This is the “holy shit” moment of AI 🤯
OpenAI just launched Sora, an AI that can create hyperrealistic videos from just text prompts.
It’ll be nearly impossible to tell the difference between AI and real in 2024.
(THREAD 🧵) 1/13 pic.twitter.com/QMlg2j2wXC
– Barsee 🐶 (@heyBarsee) February 16, 2024
Not just prompts: In addition to prompts, users can also upload still images, extend existing videos or add missing frames. The generated videos are characterized by a remarkably natural presentation, even if a certain falseness is still recognizable on closer inspection
Not perfect: OpenAI openly addresses some problems that Sora still has. Accordingly, the AI can confuse left and right, cannot correctly simulate the physics of some objects and does not correctly display some reaction sequences
A good example in which the text-to-video function worked less well can be seen here:
even the sora mistakes are mesmerizing pic.twitter.com/OvPSbaa0L9
– Charlie Holtz (@charliebholtz) February 15, 2024
OpenAI takes these challenges seriously and plans to work with a team of industry experts (Red Teamers
) to assess potential risks. Their focus is to ensure that Sora does not generate false information and is free of hate content or bias. Accordingly, a text classification is introduced to reject requests that violate the guidelines.
When can I use Sora? There is no launch date for the text-to-video AI yet. As soon as we find out something, we will let you know
Writing prompts for such AIs needs to be learned. Take a look at this article:
Now it’s your turn: What do you think of the new text-to-video AI from OpenAI? Are you looking forward to the tool or do you have concerns? What concerns would that be? Will you use Sora when it is released? And what kind of video would you create first? Could you perhaps even use Sora professionally? Let us know in the comments below:
