OpenAI introduced Sora, its artificial intelligence-powered video generation model, to media industry executives in recent weeks in an try and drum up excitement and allay concerns in regards to the technology's potential to disrupt certain sectors.
The Financial Times desired to put Sora – and his rivals – to the test.
We asked executives from promoting, animation and real estate to write down prompts to create videos that they might use of their work. We then asked them for his or her thoughts on how this technology could change their workplaces in the longer term.
Sora has yet to be released to the general public, so OpenAI tweaked a number of the prompts before sending the resulting clips. For Runway and Pika, the initial and optimized prompts were entered using essentially the most advanced models from each corporations. Here are the outcomes.
Charlotte Bunyan, co-founder of Arq, a brand promoting consultant
OpenAI's revised version of Bunyan's request to create a campaign for a “well-known high street supermarket”:
Pike and Runway's videos are based on Bunyan's original prompt:
“Sora's depiction of the people was consistent, while the actual visualization of the incredible playground was faithfully reproduced when it comes to the descriptions of the assorted elements, which others failed to realize.
“It is interesting that OpenAI modified “children” to “people” and I would really like to know why. Is it a protection issue? Is it harder to represent children because they haven't been trained in as many cases? They selected “humans” reasonably than a Caucasian man with a beard and brown hair that Sora actually spawned, which raises questions on bias.
“Pika felt surreal, such as you were in a crazy movie moment. The child version is significantly better than the surreality of the adult version, but the remaining of the environment lacks detail from the prompt. I even have a certain fondness for the liveliness of (Pika's child) version because it conveys a way of joy and happiness greater than every other.”
![A screenshot of Runway's AI-generated video showing two children on a spinning watermelon](https://www.ft.com/__origami/service/image/v2/images/raw/https%3A%2F%2Fd1e00ek4ebabms.cloudfront.net%2Fproduction%2Fc204d401-cd72-4a70-aa0d-7bd42054ca58.jpg?source=next-article&fit=scale-down&quality=highest&width=700&dpr=1)
“The runway was very much in the center. Certainly there have been fewer disruptions within the adult version, however the display of the playground elements was missing.
“I could potentially use the Sora video as a teaser for something we could bring to life in a virtual experience. It would display the playfulness of the food. However, chances are you’ll must add a human layer to the content using editing tools.
“These tools will speed up the best way we communicate creative ideas and make them more tangible. This would make it much easier for patrons, for instance, within the early stages of presenting an idea to a brand, to know what it would appear like or how it could work.
“My prompt incorporates abstract creative concepts which might be tougher for these tools. In the world of creativity, you frequently attempt to create something that didn't exist before. I do know there may be quite a lot of concern and maybe negativity about AI taking all of our jobs, but I feel we must always take into consideration how AI could make our jobs easier and ease a number of the burdens.”
Alex Williams, animator whose credits include “Who Framed Roger Rabbit?” and “The Lion King”
Videos generated by OpenAI's redesigned prompt:
“It has the slight transformation quality that AI-generated work has, which I don't think makes it client-friendly yet, but that's something that will probably be smoothed out.
“Each is amazing in what it does, but each has obvious flaws.” . . Like heads changing shape and flamingos merging with other flamingos – that doesn’t work yet.”
![Stills from an AI-generated Flamingo video from Runway](https://www.ft.com/__origami/service/image/v2/images/raw/https%3A%2F%2Fd1e00ek4ebabms.cloudfront.net%2Fproduction%2Faf1cec2b-1e28-4964-a5d0-afaf2c16c8bf.jpg?source=next-article&fit=scale-down&quality=highest&width=700&dpr=1)
![Stills from an AI-generated Flamingo video from Runway](https://www.ft.com/__origami/service/image/v2/images/raw/https%3A%2F%2Fd1e00ek4ebabms.cloudfront.net%2Fproduction%2F62fcd1f4-321c-4072-94c3-85856003dfed.jpg?source=next-article&fit=scale-down&quality=highest&width=700&dpr=1)
“I didn't manage to provide a brief film with a starting, middle and end, so it didn't do what I hoped it could. On the opposite hand, the animation could be very impressive.
“Since I began animation within the '80s, some very significant technological advances have greatly modified the medium. There is little doubt that that is the most important change I even have experienced in my profession.
“I might draw comparisons to the shift from 2D to 3D animation that was happening within the late '90s when it got here out. Initially, there was quite a lot of resistance to those changes within the hand-drawn animation community, including myself.
“It took me just a few years to understand that I needed to embrace this transformation. We all fought it together for some time, however it became a giant box office hit. As an industry we want to embrace technology since you never wish to get on the mistaken side.”
Ashley Shakibai, production manager at business real estate broker OBI Property
Videos created by the decision to create a promotional video for a business constructing in Manchester:
“Sora did an inexpensive job firstly. The transition will all the time be difficult, and that's what it struggled with. But I feel the photorealism at the tip of the shoot was quite pleasing and surprising.
“Technically the request was for people to enter the constructing, but that wasn't shown and there have been many other elements that it didn't achieve.
“All Pika got from my request is a 'sunny day.' It gave us some flares and just a few buildings, but you’ll be able to’t see the people.”
![A screenshot of an AI video created by Sora showing a couple's faces](https://www.ft.com/__origami/service/image/v2/images/raw/https%3A%2F%2Fd1e00ek4ebabms.cloudfront.net%2Fproduction%2F5e66497b-bf5c-45fe-9851-2914080e3cf7.jpg?source=next-article&fit=scale-down&quality=highest&width=700&dpr=1)
![A screenshot of an AI video created by Runway with a distorted body](https://www.ft.com/__origami/service/image/v2/images/raw/https%3A%2F%2Fd1e00ek4ebabms.cloudfront.net%2Fproduction%2Fd0557ddb-3b79-4e91-b65a-5d5bf31e8001.jpg?source=next-article&fit=scale-down&quality=highest&width=700&dpr=1)
“I needed to laugh after I saw that runway movie. It's somewhat more photorealistic, but people run forward after which backwards, so it's definitely not a believable scene.
“As an industry skilled, I expect perfection. I'm on the lookout for realistic quality videos, and AI will probably never quite have the ability to do this.
“At the tip of the Sora video, the couple is talking in a restaurant and appears like they’re having fun. This could be a chance that we’d use to sell a business property as an amenity nearby.
“At some point we’ll reach some extent where that is an incredibly powerful tool for developers and can inevitably eliminate the necessity to use other tools. Sora will seriously challenge stock web sites and the roles of actors we each use now.
“You should be very careful when adding computer-generated images. If it has no purpose, if it's not believable, it may be too distracting. It continues to be within the testing phase.”