This is OpenAI’s new video artificial intelligence

Last week it ended with one of the artificial intelligence ads of the year. OpenAI, the creators of ChatGPT, introduced Sora. It’s a powerful new video-focused model that ranges from completely creating a video with AI to editing and editing existing videos. And throughout a fast of text.

It’s something similar to the AI ​​we already have for images, but much more comprehensive. The realism that was shown in the examples it’s impressive, so much is generated doubts about this new AI. Many of them don’t have an answer, but there are many more that do.

Everything we know (and what we don’t know) about Sora and its arrival on Mac, iPhone and iPad

It is worth starting by saying that Sora will be a cross-platform model as is already the case with ChatGPT and many other generative artificial intelligence models, its arrival in the Apple ecosystem is therefore assured. However, both in relation to their operation and in others, doubts remain.

When will Sora arrive?

It’s a million dollar question and, as much as we’d love to solve it, has no official answer. What has already been shown by OpenAI shows that it is an already advanced system, although the fact of not having already announced a date suggests that there are still points to be resolved.

In fact, OpenAI itself assures that wants it to be a secure system for the user and they are working on it now. As soon as it’s ready it will be released, but we emphasize that there is no date. The assumption is that it will be launched before the end of the year, but it could well be December 31 or tomorrow (spoiler: it won’t be tomorrow).

What is Sora capable of doing?

Defining Sora as an AI capable of generating videos is telling the truth, but not the whole truth. The abilities he has are varied and very powerful:

• Videos made entirely with AI by introducing a fast in which you can either give a simple instruction and let Sora complete it at his own discretion, or request the creation of one or more specific scenes with great detail.
• Edit an existing video by ensuring that, for example, if it is a car driving on a normal road, Sora is able to give the impression that the car is driving in the middle of the jungle.
• Different video formats both for the modification of existing ones and for their complete creation. It will support vertical, horizontal and square formats in 1:1.
• Invent the environment for a video in the same way that some applications like Photoshop already do with images. For example, let’s imagine a horizontal video of a forest, because with Sora we can change it to a horizontal format in which these sides would be covered with AI-generated elements and consistent with what is shown in the original part.
• Merging two videos, which we understand better with the example provided by OpenAI: you see a video of a butterfly in flight and another video of a drone flying over a city. Well, the two videos mixed together result in this butterfly flying over the city.
• Bring static photos to life by uploading them to the system and having Sora turn them into a video. For example, a simple, normal photo of our dog can become our dog leaving that photo and walking on a beach.
• Compression of the environment and not simple video generations without more. That is, if we ask for a video showing a person walking in a city, Sora will know how to interpret the space and add more people walking in the background, vehicles on the road, buildings, etc.

Can everyone still use Sora?

Yes and no. Beyond OpenAI’s own workers (and not all of them), the company has opened a few testing limited to certain users. As you can imagine, these aren’t random users, but rather professional designers and even filmmakers, from whom OpenAI hopes to receive feedback to continue improving the model before its global launch.

How will Sora be executed?

There are up to three ways to run an AI like Sora. Locally, which implies that you limit yourself to having certain and powerful hardware on the device; on external servers, which depends only on the company that offers it and is therefore activated for many devices and finally in a mixed model which requires some execution locally with part of the process on external servers.

Also, Sora will run on OpenAI’s own servers, which ensures that it can be run on a very wide range of devices. And yes, as we have already said, also on iPhone, iPad and Mac, in fact they do not even have to be recent models. This is the same thing that is already happening with ChatGPT.

Where can you download Sora?

It’s not confirmed, but just like we have an app for ChatGPT, one could be launching for Sora as well.

This is another question that has no answer. Or at least not completely. At OpenAI they confirmed that It will be added to the catalog like the rest of the models available by the company, but not whether it will have an independent application and on which systems.

Hopefully it will be available online through a web browser. However, it is unclear if it will end up having an app for iOS, Android, macOS and Windows as is the case with other models.

What are Sora’s limits?

Without having been able to test this AI in depth, it is risky to say the least to present a list of bugs. What OpenAI has made public is that These will be videos of a maximum duration of 60 seconds duration. At least that’s what they have currently and who knows if it will be expanded in the future. For now, this minute is the limit.

Furthermore, they also comment that may have some accuracy issues when faced with complex scenes, in which Sora may not be able to understand the cause and effect relationship.

Do I have to pay to use Sora?

If the date question is the million-dollar question, it’s the nine-hundred-thousand (or maybe also a million) question. There’s no answer, although ChatGPT might be the definitive clue. OpenAI offers its chatbot for free with GPT-3.5, while the most advanced version is called ChatGPT Plus, it is based on GPT-4 and is paid.

Yet, There are several possibilities with Sora. On the one hand, it is completely free for everyone and without limits, which, although not confirmed, seems improbable to say the least. Another option is for her to be paid in full, which is more likely than the above, although it doesn’t seem like she’ll be chosen. The option that costs the most in the absence of confirmation is the mixed model with a free option accessible to all and a much more ambitious model (Sora Plus?) which does indeed require a recurring payment.

What impact will Sora have on the audiovisual industry?

Will Sora benefit or harm video publishers? (Image generated with DALL-E 3)

Well, it’s too early to know. If we’ve already seen that ChatGPT can transform our routines for the better, I hope Sora will too. However, both positive and negative consequences are expected. On the positive side, we see that it will be an extremely useful tool for video content creators and, with good evolution, even for large audiovisual producers.

But it also remains to be seen whether the famous fear of losing one’s job will eventually become a reality for professional publishers. There is also a certain risk of manipulated videos appearing and becoming fake newssomething that has already happened with static images generated with AI.

At any rate, It doesn’t seem like there’s much left to answer all the questions with certainty.. For now, Sora is looking forward to a revolution in the audiovisual field. His future is a mystery and who also knows what he will prepare for the competition. Without going any further, Tim Cook himself confirmed that Apple would soon present something regarding Apple’s artificial intelligence.

In Applesfera | An Apple co-pilot: Gurman brings iOS 18 AI to developer tools

In Applesfera | Vision Pro: price, features, release date, compatible applications and everything about Apple glasses