🎥 Open AI's Sora: First Impressions

What is Sora?

Earlier this month, OpenAI launched to the public Sora, its highly anticipated AI video generation platform.

With Sora, OpenAI has set its sights on video as the next frontier, offering users tools to generate, remix, and enhance clips using simple text prompts.

How to Get Sora

• Subscription Requirements: Sora is available exclusively to ChatGPT Plus and Pro subscribers. Plus plans cost $20/month, while the Pro plan is $200/month and offers significantly more features and capabilities for power users.

• Credits and Usage: ChatGPT Plus subscribers receive 1,000 credits per month, which translates to approximately 50 videos at 720p resolution or fewer at higher resolutions. Pro users enjoy 10x more priority credits, higher resolution options (up to 1080p), and simultaneous video generation capabilities

• Explore for Free: Even if you’re not generating videos, you can explore Sora’s Featured and Recent feeds to see what others are creating. It’s a great way to get inspired and learn how different prompts lead to varied outputs. Or, even snagging to create your own gif or meme.

Original:

Meme-ified:

Things It Can Do

Sora offers a range of tools to help users create, refine, and elevate their video projects. Here’s a concise look at some of its standout features:

• Text-to-Video: Generate creative clips directly from text prompts—perfect for artistic or abstract concepts.

• Multi-Format Options: Produce videos in various lengths, resolutions (up to 1080p), and aspect ratios tailored for different platforms.

• Remix: Tweak existing videos with simple text prompts to add new elements or modify styles.

• Blending: Seamlessly merge two videos with transitions that harmonize lighting, motion, and perspective.

• Storyboard: String together prompts to create more cohesive, multi-scene videos with a consistent style.

Things It Can’t Do (Yet)

1️⃣ Use images of people.

To address concerns about deepfakes and likeness misuse, OpenAI has restricted Sora’s ability to generate videos featuring people. Only a small group of testers can use this feature, and strict safeguards remain in place to prevent unethical use.

2️⃣ Cross the pond.

If you’re in the UK or Europe, there’s some bad news: Sora isn’t available to you yet, and there’s no clear timeline for when it will be. OpenAI’s cautious rollout echoes challenges faced by its other products, like ChatGPT and DALL-E, which have been slow to launch in regions with strict AI regulations.

3️⃣ Maintain continuity and consistency.

AI-generated videos often lack coherence across frames, especially ones involving complex actions over long durations. Characters or objects may also appear and disappear without explanation. Or, even worse—morph into some freakish twirling nightmare.

What We’ve Learned From Using It So Far

1️⃣ Be specific.

This really applies to all AI prompting—the more specific you are, the better.

For example, I wanted to see if I could get Sora to mimic a shot from our docu-series “More Than Cars.”

Here was the prompt for the clip above:

Wide shot that pulls out to get wider. Two men with beards in black, flat-brimmed hats are standing in front of a car dealership. They both have on black tshirts, jeans, and sneakers. Ends with a black screen that reads "More Than Cars" in white bold text.

For anyone who has ever met Paul or Kyle, it’s pretty clear which vague term really doomed this one. But, who knows? Maybe they’ll be inspired to try out some new headwear now.

2️⃣ The text can be gibberish.

Text has been an on-going struggle for many generative AI programs, and Sora is no exception.

It commonly leaves words out, misplaces letters, or creates some weird mashed-potato language that not even the most clever cipher couldn’t understand.

Using the original video, Sora performed a “slight remix” with the prompt:

The men should have on flat-brimmed baseball hats. Their shirts have "Love People More Than Cars" in a white font printed on them.

The hats were a total fail. The text? Close, but no cigar.

3️⃣ Even simple movements can be wonky.

Motion, especially for animals and people, is a definitely a weak point.

Using the original video, Sora performed a “strong remix” with the prompt:

The men should have on black baseball hats and be walking toward the camera as it pulls out. There should be fireworks going off and confetti falling behind them.

We finally get some hats that are much more befitting of the Automotive Troublemakers; but, did you notice how the “walking” was more of a spastic shuffle? And, as impressive as those long shadows are, do you notice how they don’t quite align with the figures?

This is all part of Sora’s movement and reality inconsistencies. As someone who failed that class in high school, Physics is impossibly hard, so I get it.

But now that Sora is being trained on a massive scale, it’s only a matter of time before it polishes those kinks, making videos even more accurate and realistic.

Have you given Sora a whirl? What did you think? Is it worth the price? Leave us a comment below! 👇