OpenAI Sora: from text prompts to high-def video

In today’s post, we’ll take a closer look at OpenAI Sora – an AI model rumored to generate videos from text prompts.

Artificial intelligence (AI) has rapidly moved from a niche technology to a major part of our everyday lives.

This post will focus on one such extension: Sora – the AI model that can reportedly generate videos from text.

Take a seat in the director’s chair

In February, AI company OpenAI (the creators of ChatGPT) unveiled a new extension to their popular AI chatbot.

This new model can reportedly generate video clips and elements based on text-based prompts.

The concept is similar to what we’ve seen with AI models for image creation, such as Midjourney5 and Adobe Photoshop FireFly AI, which allow users to create or modify images through text descriptions.

From simple prompts to realistic videos

Just as ChatGPT can provide meaningful answers to a wide range of questions, OpenAI claims that Sora will be capable of creating realistic videos from various prompts.

Sora will have knowledge about real-world details like landscape elements, realistic movements within the surrounding environment, different camera settings, video formats, and more.

I must admit, I’m genuinely impressed by OpenAI’s efforts and can’t wait to try it out when Sora is finally launched.

Sora is still in the testing phase

Since announcing Sora in February, OpenAI has been transparent about the process. It wasn’t an official launch, as Sora is still in the development phase.

OpenAI is working closely with security experts to ensure the model doesn’t produce content that violates their guidelines.

They are also collaborating with designers and video creators both internally and externally to gather valuable feedback for further development.

User involvement – part of OpenAI’s DNA

As Sora is still in the testing phase, OpenAI admits there are still errors in the video results from text-based prompts.

For instance, errors may occur in more complex prompts that require precise camera settings or involve larger scene setups over extended periods.

In my view, this is part of OpenAI’s DNA—allowing space for the public to contribute to the fine-tuning of the models and programs they release.