Sora: The Dawn of an Era in AI Generated Videos
OpenAI released Sora a few days ago, a video-generation AI model, and it has been all over the headlines ever since, taking the world by storm with its disruptive technology. It can generate videos from text prompts up to a minute long with an impressive level of visual quality.
It was only a little over a year ago that OpenAI released ChatGPT, and the world of content creation has not remained ever since. Text-generation was astounding to see AI models do with how good they were, and then image-generation came along to raise the possibilities even higher.
We had actually seen AI-generated videos before about a year ago, but they were ridiculous, unsuable, and not promising. The improvements we have now are insanely impressive. Sora's ability to generate high-quality videos while understanding the textual prompt shows the rapid progress made in AI research and development. It's an astonishing leap.
How Sora Generates Videos
The same way we generate texts and images with AI, generating videos with Sora works. A prompt describing the video is given to Sora, and then it reads it and tries to understand the details in the prompt.
Then it imagines a visual story based on the details and begins to generate frames while making sure to maintain consistency. Then it stitches all the frames together to form a complete video. At the level that Sora is at, it would only generate videos up to a minute long.
The video above was generated directly with Sora with the prompt: A movie trailer featuring the adventures of the 30-year-old spaceman wearing a red wool-knitted motorcycle helmet, a blue sky, a salt desert, cinematic style, shot on 35mm film, and vivid colours.
Where Sora Can Be Used
Unsuspecting people looking at a video generated by Sora may not think that what they are looking at isn't real. That is how good the videos can be. With that level of quality, it opens a new line of possibilities in film production, advertising, education, and gaming.
The point where AI-video generation is the lowest it will ever be is henceforth. And soon enough, it will be able to generate longer videos with even better consistency and realism. AI-generated films no longer seem far-fetched to me.
Sora has the potential to streamline content creation processes and unlock new creative possibilities. I can imagine people using it for interesting ideas like ad videos, illustrations, cinematic films from stories as prompts, and many more things that we may not even think of at this time.
This new AI model is bound to disrupt aspects of the video industry, just as the other models did to the writing and art industries, like ChatGPT-3 and DALL-E. There are implications for this new technology and the content it will create, such as video content creators feeling threatened in their jobs and reacting.
Collaboration and Feedback
Sora is not yet publicly available. "Sora is becoming available to red teamers to assess critical areas for harm or risks. We are also granting access to a number of visual artists, designers, and filmmakers to gain feedback on how to advance the model to be most helpful for creative professionals." OpenAI says in their release.
This is important for OpenAI to do for fine-tuning their new model and for ensuring that it is safe to release to the public. It's a powerful tool going into the hands of just about anybody.
Sora Is Not Perfect Yet
As much as the videos Sora generates look realistic, they are not all that perfect. If one looks closely, they'll find certain things that aren't exactly real or possible. And, as well, the model struggles to handle certain details.
The model faces challenges with properly handling complex scenes and understanding causality—such as accurately simulating the physics of a scene. For example, it once generated a video of a person running on a threadmill but facing the wrong direction.
When it comes to rendering the hands of a person, for example, it tends to do poorly with that. You could find a person's fingers looking very weird or having more than normal.
The Dawn of the AI-Video Revolution
Again, this point where we are with AI videos is the lowest it'll ever get. In a matter of weeks, months, and years, the level of performance with these models will improve tremendously.
Not only can Sora do text-to-video generation, it can also do video-to-video generation, where a video can be input and it will transform it into what's desired with the instructions it's given.
Personally, I look forward to the time when it will become available for everyone to use. There are fascinating things that will begin to show up all over the internet as people channel their creativity in a new way with this revolutionary tool.
What do you think about Sora? If you could use it, what kind of things would you generate with your creativity?
Read More:
OpenAI Bringing Common Sense to Their Models
Meet Rufus: Amazon's AI-Powered Shopping Companion
How Useful is the Most Responsible AI Chatbot, Goody-2?
AI Deception Unveiled: The Dark Side of Machine Learning
ElevenLabs Voice-Cloning AI: A Double-Edged Sword in the Voice Industry
Posted Using InLeo Alpha
Man Sam Altman is a genius imo... The way he marketed this whole thing by first asking for 5-7trillion for ai development and than launching sora couple of days later is impressive.
Only if this goes public, I guess it'll be so over for advertisement video creators...
I saw couple of trailers or game related trailer generated with Ai before sora..this is far more better.
Let's see when it'll be available for public.. won't be free afaik
They know what they're doing at OpenAI. They know just when and how to push their models so they stay on top.
I know for sure that there'll be issues with and video creators when Sora comes out. It would only be wise for them to harness it as well.
I think Sora will be free for a while or for a low power version to the public. They'd want everyone to have a taste of it, and then they'll put a price on the actual version that's better. That's what they did with ChatGPT, remember?
An amazing delivery from Open AI. Humans have a lot to learn to stay ahead.
There is so much to explre in the world of AI, and it keeps getting more fascinating.
Wow!! This is mind blowing and also scary.
So this is the future it's just too real🤯.
It really is scary. But I cannot wait to try it out when it goes public. The future is here already!
I'm looking forward to your post about it when you try it out.