OpenAI launches SORA
Day 47 / 366
OpenAI has so far come up with the GPTs which were text-based LLMs. They also came up with DALL-E, which is a text-to-image model. And just when we thought that AI developments were hitting a plateau, OpenAI came out with something yesterday that surprised everyone.
OpenAI launched SORA, a text-to-video model. The model is capable of creating 60-second-long hyper-realistic videos with just a simple text prompt.
While this model is not available for general use yet, they did provide some sample videos generated by it. You can check them out here
These videos are amazing. For instance this is a screenshot from one of the videos generated by this model.
Text-to-video is not new, there have been companies working on AI video models, but all of them felt unrealistic. They were jittery, they were more dreamlike. Which is why SORA is leagues apart from all other models.
Not everyone was excited by this. There were a lot of young startups working on AI video, using whatever open-source models they had at their disposal. They had millions of dollars in funding, but they wouldn’t have dreamt of achieving what OpenAI just did.
Of Course, we will have to wait till everyone can try the model out to see it it’s as good.