Sora's AI-generated video reaches scary levels of realism

It wasn't that long ago that we were introduced to generative AI images, made by typing prompts into programs like Dall.e, but Open AI has just introduced its video-generating program with results that may leave you questioning reality.?

Ok, perhaps that's a tad strong at this particular moment in time, but we are closer than ever to seeing realistic videos generated by a written prompt. Open AI's new program is called Sora, which is described as 'an AI model that can create realistic and imaginative scenes from text instructions'. The program is still in development and only available to a select number of people, but the videos already created are staggering and conjure questions about its future use. ?

Above: super-realistic AI video of an astronaut in a desert-like landscape

I was made aware of Sora's capabilities by Marques Brownlee's recent YouTube video, where he analyses the program and how fast video-generated AI has progressed in as little as a year. Brownlee makes the point of stating that this is the worst that this technology is ever going to be, implying that it will only become more realistic. This is a scary yet exciting notion and leaves me wondering what this means for the video and animation industries.?

Above: AI-generated video showing drone-like footage of a lighthouse on a rocky cliff

It may not be a popular opinion but I think the evidence of these videos suggests that the future of this technology will start to reduce the need to hire humans to capture scenes. I want to clarify that I don't believe that it can replace the human imagination and ability to create art, but, the example videos flying over landscapes could remove the need for drone pilots in the advertising and realty markets. I can see advertising agencies thinking - Why hire an expensive professional when you could type in a prompt with a small subscription fee? Much like many have with illustration and stock imagery.?

Above: AI generated video of woman blinking

I do think that we are still a long way off, but AI is learning faster than ever. At the minute Sora can only generate 1 minute video clips, but who knows, in the future we may see full AI-generated movies. Or if the text is all that is needed, could we start to see books being made into movies without the need for large studio budgets?

Take the above video as an example. The prompt for this was: "Extreme close up of a 24-year-old woman’s eye blinking, standing in Marrakech during magic hour, cinematic film shot in 70mm, depth of field, vivid colors, cinematic"

Above: AI generated video of an SUV on a mountain road

The videos still need a little more training and understanding of how physics, reflections, and textures interact over time to make truly life-like videos - but it's certainly getting there. The 'realistic' style videos at present remind me of a high-graphic video game, but once it figures it out, you can expect a lot more questions on safeguarding and copyrighting.

You can see more videos over on the Open AI Sora webpage. But of course, it is also to be said that the videos shown here are most likely the best of a wide selection, so take everything you see with a pinch of salt. Whether you are a fan of AI generation or not there is no denying that it is changing our world and the landscape of the imaging industries.

#ai #sora #openai #chatgpt #technology

Article source: Sora's AI-generated video reaches scary levels of realism - and this is just the start (msn.com)