OpenAI Sora Now Available for Public Use – Controversy Ensued

Artificial Intelligence

by Mascha Deikova2 days ago

After creating quite a lot of buzz back in February, OpenAI has finally released its AI video generator Sora. This tool, which can create hyperrealistic videos from text descriptions, is now available in the US and most other countries as part of the ChatGPT Plus and Pro subscriptions. What does this release include, and how are (or aren’t) developers approaching the ethical matters? Let’s find out!

In case you haven’t followed the AI news in the film industry, OpenAI announced Sora in February and instantly got a huge wave of mixed reactions. The first demonstrated tests showed an unprecedented level of realism and initiated an array of ethical questions, alongside excitement from some creators, and existential fear from others. Although Sora is not the only AI video generator on the market, as with other OpenAI products – it might become the most used one. Therefore, we need to be aware of what kind of generated content we’re going to face in the near future.

OpenAI’s Sora released – the new Sora Turbo

The newest model, which has just been moved out of the research phase, is called Sora Turbo. According to OpenAI’s announcement, it is significantly faster than the version they previewed in February. Apart from that, it includes more advanced features like animating images and being able to implement text feedback to already created video clips. In their YouTube Livestream, developers walked the audience through the tool, showcasing all the upgrades and details. You can also read about its features here.

OpenAI’s Sora is released as a standalone product on Sora.com. In case you already have a ChatGPT Plus or Pro subscription, you can try it out straight away. (Unless you’re in the UK, Switzerland, or the European Economic Area).

Sora’s interface and features

During the livestream, OpenAI demonstrated Sora’s new and simple-to-use interface, which includes all the settings and features on one screen. One such addition is called “Storyboard” – a function, that lets users generate videos based on a sequence of prompts with image consistency (and something, that made Sora stand out in its early development phase).

Apart from entering a text description, users can also choose a style preset, define the aspect ratio, increase the resolution, and tweak the amount of variations Sora should come up with. After the AI generates video clips, the software also offers further editing options: trimming the material, expanding the chosen video, blending it with another source material, and/or looping the results. However, the most impressive addition is the “Remix” tool. With it, users can give Sora feedback, using natural language, and alter the already generated content however they like.

Limitations

Until now, Sora was available only to some chosen artists and filmmakers, who shared their tests, works, and experiences. While the first published demos included only positive feedback, later on, some filmmakers admitted, that it wasn’t so easy after all getting the wanted results out of Sora, so they had to find workarounds in post-production, using traditional VFX techniques.

OpenAI isn’t hiding the fact that the newest version of Sora going out to the public also has many limitations: From generating unrealistic physics to struggling with complex actions over long durations. However, they still want to introduce this video-generation technology to the world and continue co-developing it together. In other words, we shouldn’t expect that videos created by Sora will work flawlessly from day one. Nevertheless, we should be aware, that the Internet will be likely drowned in AI-generated content over the next months (even more than it already is).

Ethical questions

So, how does OpenAI solve the question of misusing its tech for misinformation, fake news, and illegal activities? Developers mention in Livestream that videos generated by Sora will have visible watermarks and C2PA metadata, which indicate that they were made with AI. (At the same time, users who are subscribed to the Pro plan can download their results without watermarks, which sounds controversial to me). On top of that, OpenAI’s team promises quite a conservative moderation of prompts and uploaded footage:

Today, we’re blocking particularly damaging forms of abuse, such as child sexual abuse materials and sexual deepfakes. Uploads of people will be limited at launch, but we intend to roll the feature out to more users as we refine our deepfake mitigations.

A quote from the OpenAI’s announcement. More on their approach to safety and monitoring is to be found in the system card⁠

The question that still hangs in the air is what material OpenAI’s team trained their Sora Turbo on. The ethical issue with rights and licensing remains so far unresolved.

Price and availability

For users of ChatGPT Plus and Pro subscriptions, Sora is already included as part of their plan at no additional cost. However, it does have monthly limitations in how many videos and in which resolution you can create:

As announced, the new text-to-video model is available in the US and many other countries, except for “most of Europe and the UK.” OpenAI CEO Sam Altman mentions in the Livestream, that it may take “a while” before the tool is available in these areas as well.

Feature image source: OpenAI.

What are your thoughts and feelings on this news? Have you waited for OpenAI’s Sora release? Or rather, on the contrary, dreaded for this moment to come? How can OpenAI make this tool safe and welcome within our community (if this is possible)? What is it good for, in your opinion? And what negative consequences can it lead to? Let us know your opinion in the comments below! (But please, stay respectful to us and each other!)