OpenAI’s Sora Brings AI-Driven Video Creation to the Public

OpenAI shook up the entertainment world earlier this year with Sora, its text-to-video AI tool. The tool has sparked excitement over creative possibilities but also fears about job loss in Hollywood.

Until now, Sora was only available to select testers and artists. Starting Monday, it’s accessible to the public in the U.S. through ChatGPT Plus for $20 a month and can be accessed (well, in theory) at Sora.com.

In reality, as a Plus subscriber, I am still getting this message upon entering my credentials:

Subscribers should be able to generate up to 50 videos, each up to 20 seconds. ChatGPT Pro users get 10 times that, unlimited slow generations and can access longer, higher-quality videos.

OpenAI is also planning a free version and aims to roll Sora out in more countries soon.

You can check out full list of it's features and capabilies on OpenAI's website here https://openai.com/sora/

So while I wait for the heavy traffic to subside to possibly get myself in, I'm admiring the demo videos on the homepage:

In particular, the drawing video is fascinating. As the hand moves, the color appears right under the pencil, just like in real life. In other generators, the case can be that no color appears, or does so erratically, or much faster than the hand moves etc. Sora's example is great, hopefully it wasn't cherry-picked ;) Although it should be improving anyway, as Altman said this was early and will get a lot, lot better.

While I'm struggling to get into my Sora account, you can check out some hot videos already published on X: https://x.com/blizaine/status/1866208282500530503

Well, it already looks like I'm not cancelling my Chatgpt Plus subscription any time soon. I use it for programming daily, for image generation quite often, and now sure as hell if I can get 50 videos a month on top of that I'm hooked. OpenAI is becoming the new Google, your go-to spot on the internet. Email from OpenAI next?

Logged in to my Sora account, only to find this when trying to upload an image:

OpenAI seems to be turning even more strict over depiction of people. The other day Dall-E has refused to generate photorealistic random people. 

This wasn't the case just a few days ago when it did generate perfectly realistic people, I've actually been surprised because they didn't even look cartoonish like they tend to in Dall-E. Obviously, they've changed something very recently.

Last modified 16 December 2024 at 11:08