What is Sora? - OpenAI Model Explained

Robotic animal

AI is about to disrupt video creation - lowering costs, reducing time, and democratizing the ability for anyone to create amazing cinematic scenes with a prompt. I would place a bet that someone right this very moment is creating a full featured film, one prompt at a time. Perhaps we will even see the first Oscar awarded to a prompt engineer in the not to distant future.

OpenAI’s Sora is the first glance of what might be coming very soon in the next revolution of generative AI.

What is Sora?

On February 16th, 2024, OpenAI unveiled Sora. Sora is a text to video AI model. Users can write a prompt and it will create a video that tries to match the the users description. Watch OpenAI’s introduction to Sora video below.

How long are the videos?

Sora can generate videos up a minute long. In the public examples that OpenAI has shared, the length of the videos range from 8 to 59 seconds.

How can I get access to Sora?

At the time of this writing, Sora is not available to the general public. Sora is currently available to two groups of people.

  1. Red teamers, a group of people working towards improving the safety of OpenAI’s models. Applications to become a part of this group have been closed since December 1, 2023 and no information has been made available of when applications will re-open.
  2. Small group of hand picked creators, though it seems that even the likes of MrBeast does not have access even when he asked.

The best chance for you to try Sora is by tweeting your prompt the people working on Sora at OpenAI:

  1. Sam Altman: @sama
  2. Tim Brooks: @_tim_brooks
  3. Bill Peebles: @billpeeb

What types of videos can Sora generate?

The model has a deep understanding of language, enabling it to accurately interpret prompts and generate compelling characters that express vibrant emotions. Sora can also create multiple shots within a single generated video that accurately persist characters and visual style.

Source: OpenAI.com/sora

The Sora team is working to generate and simulate anything. Animals, animation, aerial drone shots, cinematic B-roll, humans, closeup shots, vehicles, and more.

What are the Sora use cases?

Social Media

At the top of the list, short-form videos on social media platforms like TikTok, Instagram Reels, and YouTube shorts will most likely explode from Sora. I predict that there will be a noticeable increase in new faceless channels being created. Sora will be a catalyst to allow anyone to start a channel while remaining anonymous.

Advertising and marketing

Imagine that you work at a travel agency in need to promote a tour package. In the past, you would have to rely on stock video or you would have to hire a videographer. WIth the advent of Sora, that video can be created with a prompt.

Sora can create unlimited unique b-rolls to help sell products and services with the benefit that you get the generated results within seconds.

Film making and story telling

I would place a wager that someone right now is using a text to video AI model to piece together a compelling full featured film 100% AI generated. If Sora can keep generated characters consistent, it will enable anyone that can tell a good story to become their own filmmakers.

Education and training

100 page employee handbooks? Encyclopedia sized technical manuals? Sora can deliver educational and/or training material through the medium of video. We all know that technical documents are not exactly page turners. Delivering dry written content through the form of entertaining content may increase retained information better.

What are the Sora video limitations?

  1. The video generated is limited to 1080p in quality, however there are many great AI upscalers such as one from Topaz Labs.
  2. Understanding of physics. The AI may have a difficulty in generating how a ball interacts with an environment in physical space.
  3. Contextual memory. If someone takes a bite out of a sandwich, the sandwich may not reflect that a bite was taken out.

What are the Sora alternatives?

While Sora is currently being developed behind the scenes there are text to video AI models that you can try today - Runway ML, invideo, and Fliki. I’m also curating a list of AI video generators here.

In conclusion

Sora appears to be a big improvement for generative text to video AI models. While OpenAI finalizes the model for public use, especially when it comes to safety, we will most likely see it first as a stand alone interface and then later combined into the ChatGPT interface similar to how Dall-E evolved.