OpenAI released its new video generation model, Sora, a few days ago. The model allows the creation of realistic videos several seconds long by inputting text. You can generate any video you’ve ever imagined within seconds using text. It is possible to create 20-second-long videos in 1080p resolution in different screen sizes: car screens, square screens, and vertical formats. The Sora interface allows you to view each frame of the generated video separately using the storyboard tool.
Sora allows access to the service with OpenAI’s Plus subscription at no cost. With the free subscription, you can generate 50 videos in low resolution (480p) and a limited number of videos below 50 in 720p resolution. The paid “Pro” subscription offers extended service options, including higher resolution and longer video durations.
To access Sora AI, you can go through the following link: https://openai.com/sora/.
By clicking on “Start Now” in the top right corner of the website, you can reach the video creation area of Sora.
The menu includes:
- Recent videos that you can view, which were created by other users.
- Liked videos that users have favorited.
- Saved videos for later viewing.
Additionally, the menu allows you to access:
- All the videos you have created.
- Your favorite videos.
You can also upload images and videos to the platform, which can serve as references alongside the text describing the video you want to create. Folders can be created to organize your generated videos.
At the bottom section of the dashboard, you can write the prompt to describe the video you want to create. In the free version of the service, you can generate videos that are 5 or 10 seconds long with resolutions of up to 720p.
There is an additional option to select one of 5 pre-generated prompts (Presets) that best describe the video you wish to create. You can also create your own Preset and share it with a friend who can use it as inspiration for their video.
Furthermore, the storyboard tool allows editing of the video and creating a new one based on the initial generated video. Parameters that can be adjusted include:
- Video resolution
- Aspect ratio
- Video duration in seconds
- Number of new videos to generate
- Defining a new Preset.
My Personal Impression:
The interface is fairly convenient and simple, though a bit slow when navigating between sections in the dashboard due to the large number of videos displayed. Video creation is relatively fast, taking around 15 to 30 seconds for a 5-second video.
The videos I created generally reflected the input I initially provided. I tried creating a funny video of the Mona Lisa typing on a laptop, but since the Mona Lisa is a well-known and fixed image, the tool took the Mona Lisa from the available images, sat her on a chair facing a laptop, but instead of typing, she just looked at the camera with her arms crossed. So, the results aren’t always perfect.
Currently, the service allows creating videos of up to 20 seconds, and no longer. In my opinion, this is more of a gimmick for now because I don’t see much use for videos of such short duration. However, as a starting point for this project, OpenAI seems to truly surprise in a good way and gives us a glimpse into what the future holds.
