OpenAI releases the latest video model Sora Turbo, which is free for members, and the website is overloaded.

The article is first published on the Intelligent Emergence public account.

Written by Tian Zhe

Edited by Su Jianxun

In the early morning of December 10, OpenAI officially released the high-end accelerated version of the video model Sora - Sora Turbo. Compared with the first-generation Sora, Sora Turbo generates videos faster.

It is reported that the first-generation Sora launched by OpenAI generates one second of video on average in 10 seconds, while in the live demonstration, Sora Turbo simultaneously generates four 10-second videos with a total time consumption of only 72 seconds.

At the same time, Sora Turbo can achieve text/image/video-to-video generation at a lower cost.

As of now, Sora Turbo has opened all functions to subscribers of OpenAI Plus and Pro without additional charges, but the usage quotas for different membership types vary:

OpenAI Plus members who pay $20 per month have a total of 50 video generation quotas per month; Pro members who pay $200 per month can generate videos at a slow speed unlimited times per month, and 500 times for accelerated video generation. If generating high-resolution videos, the available number of times will be even less.

There are also differences in video generation. The maximum resolution of the videos generated by Plus members is 720p, and the duration of a single video is 5 seconds; Pro members can simultaneously generate 5 videos with a resolution of 1080p and a duration of 20 seconds.

The OpenAI official website shows that Sora is already available in 155 countries and regions worldwide, excluding mainland China and most of Europe.

With the opening of Sora for use, the crazy influx of netizens has caused the server to be overloaded. In response, Sam Altman, the founder and CEO of OpenAI, posted that user registration has been closed, and the video generation speed will slow down for a period of time.

Source: X

An Online Video Tool Tutorial

The OpenAI team defines Sora as a creative tool that allows users to generate the desired video through a text description, an image, or a video.

They mentioned that Sora cannot generate a feature film with one click, but needs continuous optimization. In order to introduce Sora intuitively, OpenAI turned the Sora launch live broadcast into an online video tool tutorial.

If a user needs to generate a video, they need to open the Storyboard, which shows four videos to display the video details from different perspectives.

Different angles of the Storyboard display screen

In the Storyboard, after the user enters the desired video instructions in the description box and sets the style, aspect ratio, duration, number of Storyboards, and resolution, the video can be generated.

Currently, Sora can support generating videos with a maximum duration of 20 seconds and a resolution of 1080p, and the aspect ratio can be selected from three options: 16:9 / 1:1 / 9:16.

OpenAI introduced that if the user's video instruction has fewer words, Sora will fill in more details; if there are more words, it will follow the user's instructions more closely.

In the live broadcast, OpenAI entered the instruction "A yellow-tailed white crane is standing in a small stream" in the description box and placed this video clip in the front part of the timeline. Then, they entered the instruction "This crane dips its head into the water and catches a fish" for the new video and placed it in the back part of the same timeline. The two videos are not continuous, so Sora needs to generate a transition video by itself to combine the two videos into a complete video.

The results showed that Sora generated a clear video according to the instructions and created a smooth transition segment between the two video clips, making the video coherent and having a sense of story. However, no obvious fish was generated in the video, but there were splashes when the crane picked up the fish.

Two videos are combined into one video

In addition, users can also directly upload an image or video, and Sora can generate a text description of the subsequent video based on the content, and the user can freely change the instructions of the subsequent video.

For example, after submitting an image of a lighthouse, Sora will create a card to describe the subsequent video that will be generated. The user can change the instructions and adjust the position on the timeline to determine when the generated video will play.

Upload a lighthouse image, and Sora automatically describes the subsequent video

After the initial video is generated, if the user needs to optimize it, they can use the remix tool to change objects, such as replacing a mammoth with a robot, changing the expressions of characters, etc. For this purpose, Sora also sets three intensities: subtle, mild, and strong to meet the different changing needs of users.

Replace the mammoth with a mechanized mammoth

If the user is satisfied with some segments of the generated video, they can use the recut tool to edit the segments to be retained, and then expand the video through instructions to obtain a new video.

In addition, Sora also has the loop and the advanced function blend. The former allows users to make the video loop infinitely, and Sora can generate details to make the beginning and end of the video connect; the latter can integrate two completely different scenes.

Not Just a Tool, But a Path to Achieving AGI

In February this year, OpenAI first launched the first-generation Sora, which can generate high-definition videos of up to one minute based on the prompt words entered by the user. Since then, Sora has started a 10-month closed beta test and is only open to specific external personnel such as visual artists, designers, and filmmakers.

It was not until a few hours before the start of this live broadcast that the latest official video demonstration of Sora was leaked on the Internet.

During the closed beta test period of Sora, similar products in China such as Keling AI, Jimeng AI, and Hailuo AI have gained recognition from a group of users overseas.

According to the foreign website analysis tool Similarweb, the global total visits of Keling AI in November reached 9.4 million, surpassing the 7.1 million of the overseas similar product Runway; under the tweet of the leaked Sora Turbo demonstration video in advance, many foreign users said that its video effect is similar to that of similar products in China.

Sam Altman once said that the update speed of Sora is not as expected. The reason is that improving the model requires ensuring safety and expanding the computing scale.

It is reported that in order to ensure the progress of model training, OpenAI has collaborated with the semiconductor company Broadcom to develop an artificial intelligence chip for running the model, which is expected to be launched as early as 2026.

The significance of Sora for OpenAI is far more than just a video generation tool. In this live broadcast, Sam Altman emphasized that he hopes that AI can understand and generate videos to change the way people use computers, and at the same time, it will help OpenAI achieve artificial general intelligence (AGI).

However, people have different opinions on this statement. Jiang Daxin, the CEO of StepStar, once told Intelligent Emergence that he understands that OpenAI launched Sora to explore and iterate the multi-modal generation capability, so StepStar is also researching the general artificial intelligence technology along a similar path to OpenAI; Yann LeCun, the chief artificial intelligence scientist of Meta, believes that simulating the world through generating pixels is a waste of resources and is doomed to fail.

The application time of general artificial intelligence is already in the schedule of OpenAI. Sam Altman told the media last week that the first application cases of general artificial intelligence will appear as soon as 2025. People can set a very complex task, and AI will use different tools to complete it.

"The initial impact of general artificial intelligence may be small. Eventually, its influence will be stronger than people think. Just like the emergence of every major technology, a large number of jobs will be replaced."

Perhaps with the opening of Sora for use, its impact on general artificial intelligence will gradually become stronger, and OpenAI will also achieve its ultimate goal.