×
New OpenAI model accelerates media generation by 50X
Written by
Published on
Join our daily newsletter for breaking news, product launches and deals, research breakdowns, and other industry-leading AI coverage
Join Now

Breakthrough in AI-generated media speed: OpenAI researchers have developed a new model that dramatically accelerates the generation of AI-created multimedia, potentially revolutionizing real-time applications in the field.

The innovation: A new type of continuous-time consistency model (sCM) has been introduced that can generate high-quality samples in just two steps, significantly faster than traditional diffusion models.

  • The model, developed by OpenAI researchers Cheng Lu and Yang Song, increases the speed of multimedia generation by 50 times compared to traditional diffusion models.
  • Images can now be generated in nearly a tenth of a second, compared to more than 5 seconds for regular diffusion models.
  • The technology maintains comparable sample quality to traditional models despite the massive speed increase.

Technical details and performance: The sCM model offers impressive performance metrics and addresses key limitations of previous approaches.

  • OpenAI’s largest sCM model, with 1.5 billion parameters, can generate a sample in just 0.11 seconds on a single A100 GPU.
  • The model achieves a Fréchet Inception Distance (FID) score of 1.88 on ImageNet 512×512, bringing sample quality within 10% of diffusion models.
  • sCM converts noise into high-quality samples directly within one or two steps, significantly reducing computational cost and time.

Comparisons to existing technology: The new model outperforms traditional diffusion models in key areas while maintaining quality.

  • Traditional diffusion models often require dozens to hundreds of sequential steps, making them less suitable for real-time applications.
  • Previous fast-sampling methods have struggled with reduced sample quality or complex training setups, issues that sCM overcomes.
  • As both sCM and teacher diffusion models grow in size, the gap in sample quality narrows further.

Potential applications: The breakthrough opens up new possibilities for real-time generative AI across multiple domains.

  • The technology could provide the basis for a near-real-time AI image generation model from OpenAI, potentially paving the way for advanced versions of tools like DALL-E.
  • Applications in image generation, audio synthesis, and video creation could benefit from the rapid, high-quality output.
  • Industries requiring quick, high-fidelity AI-generated content may find new use cases for this technology.

Future developments: The research hints at further potential improvements and optimizations.

  • Increasing the number of sampling steps in sCM can reduce the quality difference with traditional models even more.
  • There is potential for further system optimization to accelerate performance, tailoring these models to specific industry needs.
  • The scalability of sCM models suggests that even more impressive results may be achieved as computational resources grow.

Broader implications: The development of sCM models represents a significant step forward in the field of generative AI, potentially reshaping the landscape of real-time AI applications.

  • This breakthrough could accelerate the adoption of AI-generated content in time-sensitive contexts, such as live media production or interactive experiences.
  • The improved efficiency may also contribute to reducing the environmental impact of AI model training and deployment.
  • As the technology matures, it may spark new debates about the authenticity and provenance of digital content, given the increased ease of creating high-quality, AI-generated media.
OpenAI researchers develop new model that speeds up media generation by 50X

Recent News

Apple’s AI ad draws criticism for dystopian undertones

Apple's controversial ad highlights the growing debate over AI's role in personal relationships and human interaction.

How to protect your organization from cyber threats disguised as images

Researchers demonstrate how seemingly innocent images containing hidden text can override and manipulate AI systems, exposing vulnerabilities in visual recognition technologies.

Black Friday deals on Apple Watch, AirTag and Mac

Unprecedented discounts on Apple's latest devices, including the new Apple Watch and M3 MacBooks, hint at potential shifts in consumer spending and retail strategy.