As excitement builds for the release of Sora AI, many are hyped about its potential. However, I’m personally skeptical about whether it will meet everyone’s high expectations.
Below, you’ll find a video showcasing all examples of AI-generated videos from Sora’s official releases. These videos have amazing clarity, lots of details, and look incredibly real. If you didn’t know they were made by AI, you’d probably find it hard to tell the difference from real videos.
I always believed AI would eventually reach this level, and it’s inevitable for that day to come, but I never expected it to happen so quickly – not today, not even by 2024, and certainly not in such a short timeframe.
Immediately, some wacky ideas shoot to my mind. It could be a good opportunity to make a good fortune without needing any video-making skills. I’ve always wanted to start a YouTube channel to earn passive income from ads. For instance, creating videos of funny cats and dogs, stealing content from other YouTubers and using AI to modify them, and so on.
Then, I’ve been experimenting with ChatGPT and Microsoft Copilot quite a bit lately. The text generation is as impressive as ever. However, I have to say, the image generation by DALL·E is a little disappointing (as of February 2024).
DALL·E 3 Still Sucks…
Just kidding, it doesn’t suck; it has certainly improved a lot over time. But I still find it hard to get the right images I want most of the time.
I used Microsoft Copilot to help improve my website’s design. Here are the designs it came up with. They look nice graphically, but they don’t work well for a website. They’re too busy and don’t follow modern design principles. They look more like websites from the 2000s. 🤮
Let’s give DALL·E the benefit of the doubt. Maybe it’s not great at designing webpages. I also tried asking it to make images with the description below:
“A realistic photo of 2 Shiba Inu dogs doing kung fu. Dog A is a tan Shiba Inu with no clothes. Dog B is another tan Shiba Inu with no clothes. Dog A kicks high at dog B, and dog B blocks it with two hands. The background is a park with lots of people casually walking around.”
Here are the images it generated:
The artwork is really good, and it looks great to me. But the AI didn’t quite understand the description as well as I thought it would. Why the heck is there a woman shown doing a high kick in the left photo? And why is the dog in the right photo wearing wacky pajama pants 😅? Also, Dog A isn’t even using its hands to block the kick; the AI fails to capture that action.
In my opinion, even with image generation, there’s still room for improvement before it’s perfect. And video generation requires even more work. So, personally, I don’t think Sora AI is quite ready for video production alone. However, I believe it will be a valuable tool for cutting production costs.
In the end, while it’s good to have hope for Sora AI, it’s also crucial to stay grounded in reality. Let’s not get too caught up in the excitement and make sure to maintain realistic expectations. This will give us a more balanced perspective on the true capabilities of this new technology.
Ruth Booker