It was undoubtedly (and probably will be) one of the technological news of the year given its potential, both positive and negative. The arrival of Sora, the video creation tool using artificial intelligence developed by OpenAI, has been very well received, but also criticized. Capable of converting text prompts into 60-second videos, it was only available in a trial version for a few users, but now those responsible for it confirm that will be publicly available later in 2024.
The text-to-video AI model attracted attention for its ability to create videos based on text prompts, with some examples almost indistinguishable from actual video captures. But even so, there were many unknowns about his true capabilities and, as the videos began to be analyzed, many of the praises turned into doubts.
Especially considering that the only examples we have seen came directly from OpenAI, something that will soon change. In a recent interview with The Wall Street JournalMira Murati, CTO of OpenAI, confirmed that Sora will arrive “definitely this year” and potentially in “a few months” to the public.
The interview also confirmed that OpenAI aims to allow users to “edit” the output by directly asking to fix problems in the videos. Yes indeed, the model will not work with audio at the moment.
One of the most frequently asked questions regarding Sora was related to the source of the data used by this tool. According to Murati, “publicly available” videos and licensed videos from Shuttershock were used to train the model. Yes in these videos too YouTube was included as a source is something Murati preferred not to answer.Nor are the filters that technology would have to prevent it from becoming a tool capable of violating certain laws.