OpenAI Stuns with New Generative Video Model: Sora
By Adam Pease
OpenAI Stuns with New Generative Video Model: Sora
OpenAI has recently unveiled its first generative video model, called Sora. The model stunned the Internet with its impressive abilities to generate a wide range of videos at high-fidelity. This blog discusses the news
Generative AI Is Heating Up in 2024
While the year for AI began with a relatively slow start, we are now off to the races again with Google’s recent Gemini 1.5 announcement landing on the same day as OpenAI’s impressive Sora reveal. Together, these announcements whipped up another round of hype around an emerging market that was beginning to show initial signs of disillusionment.
In OpenAI’s case, Sora leverages the same transformer architecture as GPT models to generate novel videos based on user prompts or visual inputs. While it is not yet publically available, the model appears to blow other state-of-the-art video generation models, like Runway’s Gen-2, out of the water, with its high degree of temporal consistency, long video context length, and overall lack of graphical artifacts.
What’s Next For Sora?
Sora’s announcement led to excitement amongst some and anxiety amongst others. Many in the media world decried the announcement as another sign that AI would soon be displacing creative work, while others are eager to try the model as soon as it is publicly available.
It remains to be seen though, to what extent OpenAI can realistically serve Sora to its massive and growing user base. OpenAI’s compute is split between training new models and providing existing services, such as ChatGPT, which have struggled to manage the sheer volume of users, even leading OpenAI to temporarily shut down new subscriptions. Video generation is one of the most computationally costly tasks in AI, which suggests that OpenAI make its time optimizing Sora, or find more ways to roll out the tool to smaller audiences due to the overall scarcity of compute.
Bottom Line
OpenAI has once again set the state-of-the-art in generative AI with its Sora model release. While the model is a research milestone, the costs of running and serving it to users remain unclear.
Get Ready to Transform With Us for Our First Transform Tour Stop of 2024!
Going LIVE on Thursday, March 15th, 2024 at 10 AM PT | 1 PM ET
Since the launch of OpenAI’s GPT service, organizations have been scrambling to put out their own copilot service.
In our March Transform Tour, taking place on Thursday, March 14, 2024, we will cover “The Battle of Edge vs. SaaS Computing” and “Putting CoPilots to Work in 2024 and 2025.”
Have a Comment on this?