Day 3 of the 12 Days of OpenAI event has concluded, and OpenAI has finally unveiled Sora, its highly anticipated video generation model. Here’s everything you need to know about Sora, its features, availability, and more.
What is Sora?
Sora is OpenAI’s new video generation model that creates videos from text or images. Users can generate videos up to 1080p resolution, up to 20 sec long, and in widescreen, vertical, or square aspect ratios. Unlike prior speculation, Sora is not integrated into ChatGPT but is a standalone product accessible at Sora.com. However, users with ChatGPT Plus ($20/month) or Pro ($200/month) subscriptions can access it without extra charges. Here are the features of Sora highlighted in the live event by OpenAI team:
Explore Section: Sora includes an Explore section where users can view videos generated by the community, along with the prompts used to create them. This section is great for finding inspiration of what you can generate using Sora and also improving your own prompting skills for video generation.
Generate Videos: At the bottom of the interface, there’s a text box where users can input prompts to generate videos. You can customize settings like aspect ratio, quality, and the number of videos to generate. Additionally, you can upload an image to give context to your prompt, enabling Sora to build videos that continue the aesthetics and scene depicted in the image.
Storyboard View: The standout feature of Sora is its timeline-based Storyboard View. This allows users to add prompts or images at specific points on a timeline. The generated video will follow the timeline instructions precisely. For example, if you start the timeline with an image of a dog running and add a prompt at the 5-second mark for “dog jumping a hurdle,” the video will transition seamlessly between the two, following your input.
Remix: Once a video is generated, users can tweak specific elements without starting over from scratch using the Remix feature. For example, OpenAI demonstrated how woolly mammoths in a video could be replaced seamlessly with robots.
Recut: The Recut feature lets users rearrange or edit sections of their videos, providing precise control over pacing and structure.
Loop: The Loop feature allows you to create seamless looping videos. Users have full control over where the loop starts and ends and how it transitions.
Blend: The Blend option enables users to merge two videos into one unique output. You can also control the transition speed and other parameters to achieve the desired effect.
Availability and Limitations of Sora
Sora launches globally today, except in the United Kingdom, Switzerland and the European Economic Area., where access is expected later. It will be available for ChatGPT Plus and Pro users with the following limits on video generations:
- ChatGPT Plus: You can generate up to 50 videos at 480p resolution or fewer videos at 720p each month.
- ChatGPT Pro: 500 video generations per month or unlimited slow generations. OpenAI also claims higher resolutions, and longer durations for pro version.
Currently, Sora is not included with ChatGPT Team, Enterprise, or Edu accounts and also for users under 18 years of age. OpenAI claims there will be new plans in the future for different types of users which can be available starting early next year 2025. From the looks of it, Sora might also follow Credits system similar to other image and video-generating tools. You can check how many credits will be consumed by each generation by hovering on the question mark icon.
OpenAI also introduced a faster version, Sora Turbo, that significantly reduces video generation time, producing lower-resolution outputs in under a minute—ideal for generating videos for ideas.
C2PA metadata will be added to all Sora-generated videos, so people can identify that it’s an AI-generated video from Sora. Also, there’s an internal search tool to help verify if content came from Sora. The generated videos also have a visible watermark by default.
While Sora offers groundbreaking capabilities, It often generates unrealistic physics and struggles with complex actions over long durations. In their blog, they also mentioned that the uploads of people will be limited at launch to reduce the risks of deepfakes and sexual abuse. According to OpenAI CEO Sam Altman, Sora represents an early step in video generation technology and will see significant improvements in the future.
What’s Next?
Day 3 of the 12 Days of OpenAI has raised the bar for this event. With nine more announcements to come, could we see upgrades to Sora or entirely new tools? Stay tuned as we cover every exciting reveal from OpenAI during these 12 days.
Also Read:
- Day 1 of 12 Days of OpenAI: ChatGPT Pro and o1-Full Model Launched
- Day 2 of 12 Days of OpenAI: Introduces Reinforced Fine-Tuning for o1-Mini Model
