OpenAI’s Text-to-Video AI Coming Soon, Promises Hyperrealistic Scenes

Sora, the much-awaited text-to-video generator launched by OpenAI is expected to be released to the public later this year, said Mira Murati, Chief Technical Officer of the company, in a talk with Wall Street Journal. This AI-based tool allows users to create videos that appear real-life-like, based on simple text prompts from users.

Earlier this year, OpenAI released Sora only for selected creative professionals. The current version of Sora allows users to create videos that are up to 60 seconds long.

However, OpenAI is working towards expanding the AI-based platform to create longer videos, she added. Furthermore, the new version of Sora will enable users to edit these AI-generated videos, she noted.

There are concerns that Sora could increase the risks of Deepfakes and scam videos once the technology is available to the public. Murati said all Sora-generated videos would have a watermark to enable viewers to discern between fake videos and those for genuine purposes such as entertainment or advertising.

OpenAI has tied up with stock images and photos aggregation platform, Shutterstock to provide a library. However, Murati declined to comment on issues such as the use of Sora-generated content that could potentially be used on social media platforms such as Facebook and YouTube that have so far been targeted by Deepfake creators.

Sora subscriptions are expected to be very pricey compared to those of OpenAI’s other AI tool, DALL-E, due to the enhanced features it incorporates.

One of the major issues Sora launch is likely to face is the upcoming elections in the US and other countries worldwide. In recent weeks, AI-generated fake videos, speeches and pictures showing US presidential candidates have proliferated over online platforms, especially social media.  The launch of Sora could therefore be most likely after the US elections are over.