OpenAI introduces Sora, a hyperrealistic AI video generator

OpenAI has announced public release of its hyperrealistic AI animation software Sora today – almost 10 months since it was first shown in February 2024.

In fact, OpenAI is releasing an upgraded version from the previous version: the new Sora Turbo will be available on. sora.com to ChatGPT Plus and Pro paid subscribers ($20/month or $200/month) for those in the US and most countries outside the EU and UK.

OpenAI co-founder and CEO Sam Altman made the announcement on a YouTube livestream, part of the company’s “12 Days of OpenAI” holiday announcement scheduled for 1 pm ET / 10 am PT.

Sora can create a wide range of videos from text or images, creating videos between 10 and 20 seconds long, and doing so in a variety of resolutions from 480p to 1080p, and color formats from square meters to square inches.

OpenAI created a unique new interface for the product, which includes a group or list view that the user can edit to see their genealogy.

Users can also access a feature called Storyboarding that allows them to create multiple interactive videos in the Timeline view. The model is trying to have a seamless transition between the sections – users can drag to slow down quickly or take a long and fluid time.

ChatGPT Plus users can create up to 50 videos per month at 480p resolution.

For professionals and advanced users, the Pro system offers high resolution, long duration, and unlimited generations at low speeds.

OpenAI also announced plans to release pricing models tailored to different user needs by early 2025.

Breaking news with MKBHD

The popular art of YouTuber Marques Brownlee, better known by his handle MKHBD, they spread the news about Sora’s release about an hour before.

“The rumors are true – SORA, OpenAI’s AI animation developer, is opening to the public today…” Brownlee wrote. post on social network X.

Brownlee also shared examples of videos he created using the generator / image / video-video, where he was given access as one of the twelve previous creators that OpenAI seeded before the software was released.

Brownlee shared that while Sora can produce visuals that are as clear and sometimes as good as those of newscasters or equipment analysts like himself, it also tends to show random information and signs that it was created by AI, such as vague, nonsensical text in news stories. , non-normal physics, even adding or removing random objects.

He added that OpenAI enforces strict rules against the creation of simulations of real people and violence and graphic themes.

However, in all YouTube reviewhe finally reiterated that “this is a lot for people to digest now…

It was dropped on Hugging Face in protest of the early testers

The release follows the leak of Sora on AI code to share Hugging Face and beta testers about two weeks ago in protest of OpenAI’s operation of the beta testing program. As the downloader wrote on the Hugging Face space:

“Hundreds of artists provide unpaid work through bug testing, feedback and testing services for the $150B company’s program. Although hundreds contribute for free, a select few will be selected in the competition to have their videos produced by Sora – offering a minimal compensation compared to PR and cost of the merchant that OpenAI receives.”

Sora is also entering the highly competitive field of AI video production. The way to run continues to upgrade its AI video production platform rapidly with new features including, last week, the ability to re-record dialogue in pre-existing footage and include characters’ faces. Advance AI and Chinese competition such as Kling, Hailuoand soon, Tencentboth have impressive AI video tools in the last few weeks alone.

So even if OpenAI – thanks to its success with ChatGPT and earlier, Sora’s visual display – may have a solid understanding that will help to spread the implementation of a new AI video generator to the general public, there are many competing options that seem, at least, to provide videos equal or better. This makes Sora a definite winner.