The new AI model named Sora from OpenAI can create videos up to a minute long with intricate camera movement, numerous characters with vivid emotions, and extremely detailed scenes. It can also add new content to already-existing footage or make videos from still images.
Sora functions by asking the user to provide a brief, descriptive prompt, like “A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage.” Using a vast corpus of videos that it has watched and learned from, it then decodes the prompt and mimics the motion of the real world.
Examples of SORA Videos based on Prompts
OpenAI’s Sora can generate videos up to a minute long while maintaining visual quality and adherence to the user’s prompt.
Woman walks downtown Tokyo
California during the gold rush
Tour of an art gallery
Colorful buildings in burano italy
SORA Model for Generating videos
Sora is a diffusion model, which generates a video by starting off with one that looks like static noise and gradually transforms it by removing the noise over many steps.
Sora is capable of generating entire videos all at once or extending generated videos to make them longer. By giving the model foresight of many frames at a time, the challenging problem of making sure a subject stays the same even when it goes out of view temporarily.
Similar to GPT models, Sora uses a transformer architecture, unlocking superior scaling performance.
Sora builds on past research in DALL·E and GPT models. It uses the recaptioning technique from DALL·E 3, which involves generating highly descriptive captions for the visual training data. As a result, the model is able to follow the user’s text instructions in the generated video more faithfully.
In addition to being able to generate a video solely from text instructions, the model is able to take an existing still image and generate a video from it, animating the image’s contents with accuracy and attention to small detail. The model can also take an existing video and extend it or fill in missing frames.
Sora serves as a foundation for models that can understand and simulate the real world, a capability we believe will be an important milestone for achieving AGI.
What are the challenges and limitations of Sora?
Sora is not perfect, and it still faces some challenges and limitations. Some of them are:
- Sora is not publicly available, and it is only accessible to a small group of researchers and creative professionals for feedback and testing.
- OpenAI has not announced when or how it will release Sora to the general public, or what the pricing and licensing model will be.
- The terms of service for OpenAI apply to Sora and state that it cannot be used to create content involving “extreme violence, sexual content, hateful imagery, celebrity likeness, or the IP of others.” OpenAI also keeps an eye on how Sora is used, and if it notices any misuse or violations, it has the authority to change the output or deny access.
- Sora has the potential to produce content that is unreliable, improper, or damaging. This could include facts being misrepresented, privacy being violated, or bias being encouraged.
- Additionally, Sora might produce content that is identical to reality, which could have negative ethical and societal effects like disseminating false information, inciting fear, or undermining confidence.
- Sora might not be able to respond to questions that are unclear or complicated, like those that require multiple sentences, logical thinking, or abstract concepts. Videos that need temporal continuity, causal relationships, or narrative structure may not be produced by Sora. It may also fail to produce videos that are consistent or coherent.
How to learn more about Sora?
If you are interested in learning more about Sora and seeing it in action, you can check out the following resources:
OpenAI’s blog post introducing Sora and showing some examples of its output.
Sam Altman’s tweet announcing Sora and sharing a video of a dog walking on the moon.
Sora’s website where you can sign up for early access and see more videos created by Sora.
Sora’s YouTube channel where you can watch more videos generated by Sora and subscribe for updates.
Sora’s Instagram account where you can see more images and videos created by Sora and follow for more content.