SORA has garnered attention for its ability to generate videos from text inputs. This article will set aside its capabilities to focus on its limitations and flaws.
SORA’s Limitations
1. Inaccurate Physical Interactions
SORA may struggle to accurately replicate the physics of everyday interactions, even simple ones.
For example, when a glass falls and shatters, replicating this event in a video may not be straightforward, and SORA might not represent it accurately.
Similarly, interactions involving food, such as taking a bite out of a cookie and leaving a mark, may not be precisely depicted by SORA. Take a look at research paper for more info.
2. Spontaneous Object Appearance
Videos generated by SORA may sometimes include objects that appear spontaneously, irrelevant to the context or subject.
This can be considered a flaw, as it may also lead to inconsistencies in longer videos.
3. Restrictions on Sexual Content
You cannot generate videos containing sexual content with SORA. Learn more about this.
Due to OpenAI’s safety measures and precautions, any text input requesting the production of such content will be rejected.
4. Prohibition on Celebrity Likeness
Attempting to create a character resembling a celebrity will also be met with rejection due to OpenAI’s safety measures.
SORA will not process text inputs aimed at generating celebrity likenesses or infringing on the intellectual property of others.
5. Video Length Limitation
According to OpenAI’s announcement, SORA can currently produce videos with a maximum length of one minute. This is a significant limitation for those looking to create longer content.
6. Resolution Cap
Videos produced by SORA are limited to 1080p quality.
If this resolution does not meet your needs, it’s important to note that SORA cannot generate videos in higher resolutions at this time.