Just a few years ago, AI-generated video was a flashy research topic for amateurs, but just this week OpenAI’s Sams Altman and team released SORA, which is making bleeding edge video in machine learning, photorealistic. Right Now, It is currently in limited access for researchers, visual artists and film-makers to test out and provide feedback.
Contents
Introduction
SORA AI is a new tool that uses the underlying AI model developed by OpenAI to create videos from text descriptions that smoking can “generate up to 60 seconds of video with highly detailed scenes, complex camera motion, and multiple characters with vibrant emotions”. SORA AI’s ability to pose characters and create animations in perfect synchronization with the motion of the camera would be impossible to create with today’s classic CGI techniques. Unlike these methods that create a nice picture and then wiggle the parts, SORA AI makes a deep neural network learn to understand and represent the real world in motion.
Features
SORA represents a significant leap in video design because it explains the process with examples. In this case, SORA shows how to create a high-quality video where it adjusts these parameters:
Cut duration and scenery complexity by 80%, helping to generate high-quality video in minutes instead of hours
Symmetry break: By making the past and future scenery different, the tool can generate high-quality creative videos that transcend boundaries
Physical understanding with AI: Beyond changes for effects and functionality, SORA uses AI to understand and simulate the world’s physical motion
Interaction: The system presents successes, but it also introduces a way for creators to give suggestions when it makes mistakes.
Advancements in AI-generated Videos
It’s a capability that almost sounds like science fiction — the ability to generate video just by writing what you want.
Ladies and Gentleman, SORA is an ideal solution for those of you who’ve had nightmares about the stop-motioning Momo on Twitch or YouTube. Watching these kinds of videos, you can definitely tell that something is off. If you knew what to look for, then it will be clear that the hands of the player in the game are standing on end. There are also points where objects seem to jump around all over the screen. As time goes on and the software advances this will happen less and less but for now, it’s a minor distraction to an otherwise mesmerizing demo.
Implications of AI-generated Videos
I’ve worried that this would create mass deception in the video age, especially in discussions of fake politicians, fake fast food owners, and fake eager applicants for a certain restaurant franchise. As you can see, there is certainly the potential for misleading a viewer, particularly in an age where many of us are hesitant to believe what we see and hear. There’s a point at which these digitizations become so good that we’re not entirely sure they aren’t real.
The flip side to this is the reality that stock footage and even some advertising could be made so quickly and cheaply that it barely required any effort
Basically, the new techs could replace a lot of traditional videography and photography. Why go to the trouble of a real shoot if the fake is easier and cheaper?
AI Generated Videos Examples
| Example | Description | Observations |
|---|---|---|
| Tokyo Street | Stylish woman walking down a Tokyo street with neon signs. | Accurate lighting, materials, reflections, some imperfections in movements and frame rates. |
| Vintage SUV | Camera following a vintage SUV on a dirt road. | Solid footage, slightly video game-like appearance. |
| Golden Retriever Puppies | Puppies playing in the snow. | Impressive physics simulation, realistic fur and movements. |
| Young Man on Cloud | Young man sitting on a cloud, reading a book. | Realistic textures, lighting, some imperfections in motion. |
| Space Adventurer Trailer | Trailer featuring a spaceman in various scenes. | Convincing visuals, cinematic style, some minor physics issues. |
Challenges and Considerations – SORA
Many companies have submitted papers or created products on AI-generated videos, but most have had access to extremely limited source material. At the same time, there’s the ever-present concern about deepfakes and other face swaps being used for harassment and other harmful purposes. There’s less chance of AI creating a harmless but unauthorized likeness of a celebrity than there is of someone creating scurrilous fakes. Nonetheless, the mere act of creating such easy-impression videos raises ethical questions and could ultimately raise legal issues. Today, it may be more important to lay the groundwork for responsible use than to tackle the technical challenges of AI-generated video.
Implications and Analysis
| Aspect | Description | Implications |
|---|---|---|
| Stock Footage | AI-generated videos can pass for real footage, useful for stock libraries. | Reduces the need for hiring photographers/videographers, lowers costs. |
| Realism | AI-generated videos are approaching realistic quality, with some imperfections. | Potential to deceive viewers, especially in sensitive contexts like elections. |
| Safety Concerns | Need for safeguards against misuse, such as preventing creation of misleading content. | Ethical considerations in content creation and dissemination. |
| Future Potential | Speculation on future capabilities and impact. | Questions on creativity and innovation vs. replication of existing content. |
Conclusion
Introducing SORA : Sam Altman, OpenAI, and others built a model called SORA that generates a one-minute video clip from text inputs.AI-generated videos have come a long way in a short amount of time. This new technology has the potential to be both impressive and useful, but it also raises concerns about deepfakes and the future of stock video generation. While it’s important to consider the potential risks of SORA and other AI, it’s worth remembering they’re still very much a work in progress.