ChaIos Media News - The Sora Event

Hi there, fellow generative AI enthusiasts. The past week has been the craziest of the year so far. OpenAI has rocked the scene by unveiling their new ‘Sora’ video model. Ever since the first footage surfaced on social media it has been on everyone’s mind. It stirred enthusiasm but also a tornado of fear and anger among those who are less enthusiastic about AI. To us one thing became clear: AI capabilities are not improving in a linear fashion. The progress is exponential. Our minds have been blown. And it’s only February. Let’s dig right in and start with the special edition of our new TV format we have recorded on the day Sora was first announced:

OpenAI unveils ‘Sora’, a video model

Like we said in our intro, OpenAI had unveiled but not yet release a model that has rocked the generative AI scene. The quality of the videos produced by this model isn’t just a little better than what we have been used to so far. It’s a massive leap forward that has us all wondering what else OpenAI may be hiding in their offices and on their GPU clusters. The ‘Sora’ model is capable of generating videos up to a whopping 60 seconds in length in a quality that puts everything else currently on the market to shame.

At the moment we do not know much else besides the fact that Sora has extremely promising capabilities. OpenAI has not provided us with an estimate of when we can expect to have access to this model or how much it will cost subscribers to use it. All we know at this point is that OpenAI is allowing a group of people to test the model in order to build robust abuse prevention. We hope this will not result in overzealous censorship that will unnecessarily cripple the model. With all that said: exciting times are ahead. AI is advancing at breakneck speed and we are excited about where this will take us.

Read more and watch demo videos at OpenAI

RunwayML announces winners of the gen:48 AI cinema competition

Two weeks ago we were completely engulfed by the RunwayML gen:48 AI video competition. Hundreds of amazing productions were submitted but only one entry could win the ‘Grand Prix’ and that honor went to the short film titled ‘Ellie’ by Unveil. An interesting and truly original short story about a sentient main character in a video game. Congratulations to the Unveil team!

Google releases Gemini 1.5 with an astonishing 1M token context window

While Sora may have stolen the show with all the amazing videos doing the rounds on social media, Google has impressed many with their Gemini 1.5 LLM. The model takes yet another massive leap forward by introducing a massive context window of one million tokens. This is enough to contain an entire book. A lengthy one at that. While not as dazzling as the videos from Sora we realize this is a true game changer with massive implications. Unfortunately we do not yet have access to this model so we are unable to give you an in-depth review but initial signals from people who do have access are extremely positive and encouraging sounding.

If any of our readers happens to be able to grant us access to this model please don’t hesitate to contact us.

Read more about Gemini on the Google Blog

FinalFrame.net hit by anti-AI protesters

Our friends over at FinalFrame.net had a rough couple of days after openAI’s Sora was unveiled. Angry AI haters attacked the small business racking up GPU cost and causing outages for them. This has cost founder Benjamin De Kraker quite some money and probably some sleepless nights which is why we are encouraging you all to buy some credits with them in support. Final Frame is a really neat AI video service powered by Stable Video Diffusion.

Watch this video of founder Benjamin De Kraker explaining what happened to Final Frame this weekend.

And that’s a wrap

That’s it for this week. This newletter is shorter than usual but really, all the generative AI buzz was dominated by OpenAI and a little bit by Google. These two behemoths literally sucked all of the oxygen out of the room this week. We suppose in a good way even though it’s unknown when the general public will actually get access to these exciting new models.

Thank you for your support and see you again next week! Got a tip, want to be featured, interviewed, or have something you think should make it into next week's edition? Reach out to Marco or Ben on X, or drop us an email at [email protected]om. Your input is always welcome!