OpenAI's Sora is making waves as a new AI video generation tool, and it's generating both excitement and concern. It was a long wait for it to be available to the public, but I manage to sign up under the ChatGPT plus plan for $20. Spoiler alert there are serious limitations to this plan with 1000 credits like 5 second videos at 720p and watermarking. I used up my credits in one day. The pro plan improves on these limitations at a cost of $200!
Here are some samples of my Sora videos
I was surprised that it knew what a bmw z4 was. It did a pretty good job
This technology allows users to create videos simply by typing in text prompts. Here’s a breakdown of what Sora can do, its limitations, and what it means for the future of online content.
How Sora Works
Sora uses a text box interface where users can type in prompts to generate videos.
The tool provides a library of videos that users have created with prompts.
Users can see the prompts that others have used to generate their videos, which can be used for inspiration and learning.
Sora also offers a remix feature that allows users to modify existing videos with new prompts, controlling how much the new video will change from the original.
The platform also includes a storyboard feature to string multiple prompts together to create longer videos.
Strengths of Sora
Sora can create abstract videos with various textures, colors, and gradients that could be used for screensavers or backgrounds.
It is capable of reproducing text, particularly when asked for specific text, making it useful for title slides and introductions.
Sora excels at generating cartoon-style videos, where its limitations in physics and movement are less noticeable.
It can create impressive fluid dynamics, such as water rippling or crashing waves and fire.
Limitations of Sora
Object Permanence: Sora struggles with object permanence; objects may disappear, reappear, or move in illogical ways.
Physics: The tool has difficulty with realistic movements and physics, often making motions appear slow or unnatural.
Leg Movements: Sora commonly makes errors when animating characters with legs, often switching the front and back legs while the character walks.
Copyright Restrictions: Sora is designed to avoid generating copyrighted material, public figures, or logos and will not create images of anyone it thinks is under 18.
Image to Video Issues: When using uploaded images to generate videos, Sora can produce "wonky" results, due to the AI not understanding the context of the direction of any object in the image
The Potential Uses of Sora
Abstract backgrounds and screensavers
Title slides and documentary introductions.
Cartoon-style videos and social media content.
Concerns and Ethical Considerations
Sora's ability to create realistic-looking videos raises concerns about the potential for misinformation and the difficulty of distinguishing between real and AI-generated content.
The tool watermarks its videos, but these can be cropped.
There are concerns about the source material used to train Sora and whether it includes private or copyrighted content.
The amount of energy the tool uses is not clear.
The tool is so powerful that it is moving us further into an era of not being able to believe what we see online.
It's a potential disrupter to the animation and stock footage industry, but in my testing experience it's still early days. It is still too costly for full HD & 4K videos and needs more improvements. For trending videos like pickleball I can tell it hasn't been trained on enough recent data as it doesn't display paddles.
Sora Reviews One reviewer had access to the tool and was able to experiment with it for several days, generating photorealistic and cartoon content with a variety of subjects. The reviewer noted both the strengths and weaknesses of the platform, highlighting issues with object permanence and physics, but also its capacity to produce impressive animations and abstract designs.
Conclusion Sora is a powerful tool that is about to be in the hands of many people. It is a groundbreaking tool for AI video generation, but it comes with significant limitations and ethical considerations. While it has great potential for creative applications, it also poses a challenge to the public’s trust in online content.
To learn more about Generative AI with step by step working examples and guidance including using AI to create text, audio, videos, and music check out our online course.
Komentarze