This blog was last modified 444 days before.

What is Sora

Sora, which is a completely new large-scale AI model published by OpenAI, gains ability to generate videos based on text prompts.

Check Open Sora Introduction Page for more info.

Sora has been created based on deeper research of previous large models like GPT and DALL-E etc. They use small video clip to train the model, which play the role like token in the training process of GPT model.

What Can Sora Do

• Generate videos based on text prompt.
• Extend generated videos and adding more details on it.
• Generate videos based on given static pictures.

Here we recommned you check out the examples privided by OpenAI in its offical introduction page.

However this new model also has limitations and drawbacks, such as the lackness of the ablitiy regarding spetial awareness and the difficulty of complying to the real world physics.

Can I use it now?

For most of the people, NO. OpenAI only open this new model to few people such as some vision artist and movie director in order to collect the opinion and feedback from related industry, which will be used to optimize this model.

Sora is currently under fine-tuning process, and also be subject to several security research and testing which aim to prevent the content generated by this model being mis-used.

Also, OpenAI claims that they are working on a detecter which can tell when a video is generated by Sora as a way to prevent such model being used to generate misleading content.