On Saturday, in a not-so-popular cryptic post on X, OpenAI chief Sam Altman hinted at Sora’s release this week. “I am so, so excited for what we have to launch on day 3. Monday feels so far away,” he said.
Needless to say, OpenAI’s popular text-to-video tool, Sora, has once again become the talk of the town after the leaked footage and information hinting at the video generation quality of Sora v2 surfaced.
Ruud van der Linden, CEO at Lont, uploaded the video of OpenAI’s Chad Nelson presenting Sora v2 at the C21Media Keynote in London.
Sora v2 release is impending:
* 1-minute video outputs
* text-to-video
* text+image-to-video
* text+video-to-video
OpenAI's Chad Nelson showed this at the C21Media Keynote in London. And he said we will see it very very soon, as @sama has foreshadowed. pic.twitter.com/xZiDaydoDV— Ruud van der Linden (@RuudNL) December 7, 2024
Sora v2 is expected to generate videos up to one minute in length, making it a significant leap from previous versions or competitors that often offer shorter clips.
This latest development offers a flexible approach to video generation ranging from text-to-video, text+image-to-video, and text+video-to-video. This could drastically change the process of video generation and interactivity and customisation, which is not commonly seen in current models.
Recently, Sora’s API, available to some artists for early testing on Hugging Face, was also leaked. Regardless of the hype that the leak created, it brought out a bigger question: Was the Sora leak real?
Not long after the tool’s leak, the Hugging Face page seemed to be failing with a 502 error due to high traffic. The company learned of this incident soon enough and shut down the access three hours after the revelation.
According to AIM, three hours seem to be good enough for users to produce content and create hype but not too long for the situation to get out of control.
Meanwhile, Google Cloud introduced Veo, a video generation model, and Imagen 3, an advanced image generation tool, on its Vertex AI platform. Veo, currently in private preview, generates high-quality videos from text or image prompts, enabling businesses to create realistic and coherent footage efficiently while reducing production time and costs.
Sora is Going to Soar High in Video Quality
Ashutosh Shrivastava, an AI enthusiast, took to X to express his thoughts on this. In the past year, while Sora has been kept from public use, many other models have come up that can “create videos as good as, or even better than, Sora”.
However, with the latest video leak, others have also speculated that Sora might surpass all other tools in quality benchmarks.
Alex Volkov, the host of the ThursdAI pod, wrote on X: “I take back EVERYTHING I said about other video models catching up to SORA even remotely.” He also hints at the OpenAI $200/mo pro tier, saying that if Sora is added to that, the company will see a lot of new subscriptions.
If the tool is not released in the ‘12 days of OpenAI’, which many are speculating that it will, OpenAI could lose its advantage with the tool’s popularity on the rise currently. The tool was then soaring the charts, also reported by AIM in June this year, expressing that ‘OpenAI should release Sora before it’s too late’.
Multiple users online have been discussing the videos generated by Sora since the API was leaked. When looked at critically earlier, the videos do not stand out enough to call this a ‘leak’ in the first place.
A user on X expressed the same: “It actually was a bit underwhelming, tbh, at least if the leaked videos are legit.”
“We have seen better now… So yeah,” the discussion continued.
The users hoped for Sora to have moved forward to a new level, which could be making a cinema-length movie out of many snippets cut together.
Another contributor expressed, “The videos it generated are quite poor compared to Kling’s.” However, quite a few were impressed with the quality of the videos generated. A user said that though “Sora is not perfect, the video is a lot more coherent than Runway.”
Following the leak, many content creators and enthusiasts admitted they’d lost all hope of Sora ever being released to the public. While early testers are still finding issues that need to be resolved before public release, a user feels that there aren’t enough resources to support a large number of users.
She also believes that OpenAI plans to offer the tool only to the biggest paying customers (enterprise users, Hollywood, etc). Some believe that the company wanted to “improve continuously as ideas came in from other video generation tools, and release Sora only when they managed to create something competitive”.
This could take a sharp turn if OpenAI releases Sora on Monday, as many speculate. An AI enthusiast on X said, “Sora looks promising and stunning! Sora might get released in OpenAI’s 12-Day campaign.”
Sora Hasn’t Died Yet
As speculations around the Sora release rise, there is considerable excitement regarding its potential public release. Some indicate that it might be part of OpenAI’s upcoming initiatives or campaigns.
With multiple scenes and character consistency across the video duration, there has risen significant interest and anticipation among tech enthusiasts and professionals in the field.
As OpenAI engaged with Hollywood through Sora, Runway partnered with top entertainment and media like Lionsgate to develop customised versions of Gen-3 Alpha. Unlike OpenAI, Runway has also made Gen-3 Alpha available to all users, though the model remains subscription-based.
With the rise and impact of other tools like Runway, Midjourney and KlingAI over the past year, it has become difficult for creators to think back to the capabilities of Sora. AIM had previously compared the launch of OpenAI’s Sora to a ‘ChatGPT moment in video generation’.
The post OpenAI’s Sora is So Not Dead appeared first on Analytics India Magazine.
