OpenAI previewed Sora back in February, leading to an agonizing wait for enthusiasts around the world. That wait has finally come to an end as Sora is now available as part of OpenAI's 12 days of shipping products which started with a ChatGPT Pro subscription and the full release of the o1 model.

The text-to-video model stunned the world when it was first previewed but since then, it was only available to select artists, filmmakers, and a select group of testers. With Sora, users can create videos using text prompts, images, and videos. The model is capable of creating complex scenes with multiple characters, and different types of motions. OpenAI says that the model is not only capable of understanding the user's request from the prompt but that it also has an understanding of how those elements exist in the physical world.

The video model is available to ChatGPT subscribers (Plus and Pro) in the US and most other countries. However, in most of Europe and the UK, the model won't be available right now, and it might be a while before it becomes available. It's probably because of the stricter digital privacy laws in the UK and the EU which OpenAI is still working through.

ChatGPT Plus subscribers ($20/month) can generate up to 50 priority videos (1000 credits) up to 5 seconds long with a maximum 720p resolution.

ChatGPT Pro subscribers ($200/ month), on the other hand, get up to 500 priority videos (10000 credits), unlimited relaxed videos, longer video generation up to 20 seconds, a maximum resolution of 20 seconds, and the ability to download videos without watermark. The company also plans to make different types of pricing available for different types of users down the line.

Sora comes with a bunch of features, like Remix which allows you to change certain parts of the video through text prompts, Recut for extending any isolated scenes from a video, Style Presets, Loop, and Blend to seamlessly combine two videos.

The interface also includes a "Storyboard" where you can generate a sequence of videos using a series of text prompts.

As expected, the company is treading lightly and is adopting a conversation approach with its video model. All videos will have a watermark (Pro subscribers can remove it, though) and will be generated with C2PA metadata which will make it easier to identify its origins as Sora. They've also built an internal search tool that uses technical attributes of generations to identify whether they've been generated using Sora. It's an important step to prevent misinformation, scams and deepfakes.

The company also requires users to check off an agreement before uploading any images and videos to confirm that they have the rights to use the video and that the video doesn't contain any people under 18 as well as any violent or explicit content. General uploads of people will also be limited at launch and the feature will only be expanded as OpenAI refines their deepfake mitigations.

Sora is available independently of ChatGPT through a new interface at sora.com. Users can enter their prompts from the bottom of the screen and configure various settings for preset styles, duration, resolution, aspect ratio, number of variations, etc. before generating the video. The interface also shows how many credits the generation will use up with those settings. Sora generates 4 variations of the prompt by default.

However, due to heavy demand, it might not be possible for you to access Sora as the creation of a new account has been suspended. Sam Altman says they are working on resolving the issue.

If you're unable to access Sora or are not a paying subscriber, you can still explore the feed of videos generated by other people.

The video model has a few limitations currently. It can often generate unrealistic videos and struggles with complex actions in longer videos. However, as the model evolves, it'll be interesting to see how it shapes the future of video creation and consumption.