OpenAI has recently introduced a fresh AI model Sora. This Text-to -video system has the capability to produce videos spanning up to one minute in duration.
Earlier this year, Bill Gates from Microsoft and Sam Altman from OpenAI chatted on the Unconfuse Me podcast. They talked about lots of stuff related to making ChatGPT, including how Altman got fired and then hired again, how they are working on GPT-5, and even super-smart AI.
Altman from OpenAI also talked about how lots of people want the chatbot to show videos, so they are thinking about adding that feature. He mentioned they are already working on letting the chatbot talk and show pictures, and adding videos would make it even cooler.
Just a month after letting people know about it, OpenAI has come out with a new thing called Sora. It’s a special computer program that can take words and turn them into video that last up to a minute. And what’s cool is, it can make sure the videos look good and match what the user wants them to do.
Finally, OpenAI has made its own video-making tool because other companies like Stability Ai and Google already have one. It seems like Open AI wanted to make sure their tool was really good before showing it off. If things keep going like this, their videos might be even better than the ones from other companies.
According to the official page, Sora can make realistic and creative scenes from just one text command, like other tools that turn words into videos. But what makes this one different is the cool technology behind it.
Making Realistic Content
Open AI says its and AI that can understand how people and things are in the real world. That means Sora can create scenes with lots of different things happening, like people moving around, showing different emotions on their faces, and lots of different objects with lots of details. The videos it makes do not look fake or weird like some other AI videos do – well, mostly.
OpenAI Sora has another cool feature: it’s multimodal, means that it can do lots of different stuff. If you have a picture, you can upload it m and Sora will turn it into a video. It’s not just any videos though-everything in the picture will start moving, and also it pay attention to all the tiny details to make it look realistic. If you have a video that’s missing some parts or you want to make it longer, Sora can help with that too! It will fill the missing bits or add more detail to it, so you can have better and complete video.
Prompt: A litter of golden retriever puppies playing in snow. Their heads pop out of snow, covered in pic.
You can see some video clips on OpenAI website and on X (Twitter). One of our favorite show’s puppies playing in the snow. If you look closely, you will notice that their fur and snow on their nooses looks very real. Another good clip shows a Victoria-crowned pigeon moving around like a bird.
Still Working on it
Even though these two videos are cool, Sora is not perfect. OpenAI Says its “model has weaknesses.” Sometimes it can’t get how things move right, mixes up left and right, and doesn’t understand causes and effects. Like, you can make a character in the video bite a cookie, but the cookie won’t look bitten.
It messes up in some funny ways too. Like there’s is a video where some archaeologists dig up a big piece of paper, and then it turns into a chair and ends up as a crumpled piece of plastic. Also, sometimes it doesn’t spell words in right way. It writes “Otter” as “Oter” and “Land Rover” as “Danover.”
In the future, the company plans to collaborate closely with its “red teamers,” team of experts in the field, to thoroughly examine various aspects of Sora’s operations. Their primary goal is to ensure that Sora doesn’t inadvertently produce incorrect information. Spread hurtful content, or display any form of bias. Additionally, OpenAI is taking proactive steps by implementing a text classifier tool. This tool will be used to identify and reject any prompts or inputs that go against their policies. Such prohibited content includes requests for s-xual content, violent imagery, or discussions involving celebrities in inappropriate contexts, among other things. By employing these measures, the company aims to uphold high standards of accuracy, fairness, and safety in Sora’s outputs.
There’s currently no information available about the official launch date of OpenAI Sora. We have connected for details regarding the release but haven’t received any updates yet.
