OpenAI's Sora text-to-video generator will release for public this year
What's the story
OpenAI, a leading artificial intelligence research lab, has announced plans to make its innovative text-to-video generator, Sora, available to the public in late 2024.
The announcement was made by OpenAI's Chief Technology Officer Mira Murati in an interview with The Wall Street Journal.
Initially introduced in February, Sora is capable of creating hyperrealistic scenes from text prompts.
As of now, this tool is exclusively accessible to visual artists, designers, and filmmakers.
Capabilities
Sora's capabilities and future improvements
Sora is capable of generating videos up to 60 seconds long, a significant improvement over its industry counterparts.
The time taken to generate these videos depends on the complexity of the prompt.
OpenAI is also working on optimizing Sora's generation time and cost-effectiveness, as well as minimizing imperfections in the model.
Murati said the company is striving to make this technology a tool that people can edit and create with, indicating future enhancements for users.
Sora's data training
Sora's training data and misinformation concerns
When asked about the data used to train Sora, Murati did not provide specific details but confirmed that it was trained on publicly available data, including content from Shutterstock.
As the 2024 presidential election approaches in the US, concerns about AI tools creating misinformation have increased.
In response to these concerns, Murati stated that OpenAI will not release anything they don't feel confident about in terms of its potential impact on global elections.
Expected enhancements
Future enhancements and pricing for Sora
OpenAI has plans to enhance Sora's capabilities by incorporating audio, which could make the generated scenes more realistic.
The company is also working on allowing users to edit videos produced by Sora.
Despite being a more expensive model to run when compared to other OpenAI tools, Murati stated that they aim to make it available at similar costs as DALL-E, their AI text-to-image model.
Insights
Potential for AI-generated adult content
OpenAI has not ruled out the possibility of Sora being used to generate adult content.
Murati mentioned that they are working with creators to determine the level of flexibility the tool should provide.
However, she also acknowledged the potential dangers of such technology, citing recent incidents involving AI-generated deepfakes.
Despite these concerns, OpenAI believes that realistic AI-generated videos could potentially revolutionize industries like pornography by addressing issues around sex trafficking and abuse.