You can now experience Sora, OpenAI’s AI video generator.

You can now experience Sora, OpenAI’s AI video generator.

Sora, OpenAI’s AI video generator, is now widely accessible. Initially announced earlier this year, the tool demonstrated the ability to generate photorealistic short videos, such as puppies playing in snow or drone shots of cityscapes. While the demos were impressive, they also raised concerns about the potential for misuse. Until now, Sora had not been available for public testing, but this week, OpenAI made it accessible to anyone with a ChatGPT Plus account.

Sora allows users to generate videos based on prompts, with the latest features adding a variety of video editing tools. “Storyboard” enables users to combine different video prompts into a continuous scene. For example, you could prompt Sora to generate a crane standing in water, and then another prompt showing the crane dipping its head. “Recut” isolates parts of a video to extend specific segments, while “Remix” lets users tweak existing videos by specifying changes, with a “strength” setting to adjust how much is altered. “Blend” allows for transforming one subject into another, such as turning a butterfly into an orchid.

When you submit a prompt, the video is added to a queue, and its generation time depends on the resolution, duration, and variations you request. ChatGPT Plus users can create videos up to 720p resolution and five seconds long, while ChatGPT Pro users can generate videos up to 1080p resolution and 20 seconds long, with more variations. However, due to high demand, some users have experienced long wait times, with OpenAI temporarily halting account creation.

While the quality of Sora-generated videos can be photorealistic, the movement often lacks consistency. In a test video of a taxi driving through a city, the taxi’s movements were erratic, and the scene quickly became unrealistic. OpenAI’s Marques Brownlee also noted that Sora struggles with realistic movement, especially in walk cycles or when objects suddenly disappear. However, motion graphics and animated clips appear to fare better. Drone shots and tracking shots are also impressive, with some minor glitches that are difficult to notice in less demanding contexts.

The concern with Sora’s accessibility is its potential for misuse. While videos may not always be perfect, they can still deceive viewers, especially with the ability to generate multiple videos quickly. For $20 per month with a ChatGPT Plus account, users can create up to 50 five-second videos a month. ChatGPT Pro users, at $200 per month, can make up to 500 videos, with longer durations and no watermarks, making it easier to pass off the videos as genuine.

OpenAI has implemented safeguards, such as blocking the inclusion of copyrighted materials or famous figures in videos, but the potential for spreading misinformation remains. If you’re interested in trying out Sora, you’ll need a ChatGPT Plus or Pro account. Currently, account creation is paused, but this may change soon. Visit sora.com to log in and start using the tool.

Show 2 Comments

2 Comments

Leave a Reply

Your email address will not be published. Required fields are marked *