Sora is impressive—until it completely forgets what it was doing. I’ve played with it enough to know where the seams are, and they’re not small.
1
No Long-Form Videos
My biggest gripe with Sora is that you can only make 5-20 seconds long videos. At best, this could be a short-form video or B-roll. It’s hard to be creative within such a short period, and at best, I can see the tool being used for experimentation only.
Since I have ChatGPT Plus, I’m limited to 10-second videos at most. I don’t think that the $200 per month fee for ChatGPT Pro is worth paying to make 20-second videos, either. If it was top-of-the-range, I could maybe see the value—but in its current form, I don’t see many benefits.
2
The “Subtle” Remix Feature Does Very Little
There’s a difference between subtle editing and no edits at all. Unfortunately, I feel like Sora’s dedicated tool for this falls into the latter category.
Each time I’ve used Sora’s subtle remix feature, I’ve been more and more tempted to switch entirely to a Sora alternative. I haven’t noticed any differences between the videos I’ve uploaded and asked to edit, other than that the image quality is worse than before.
I expect that OpenAI will fix this problem as Sora becomes more established. However, the feature’s early version feels unusable.
3
Watermarked Videos
I know that I could remove the watermark by upgrading to ChatGPT Pro, but I don’t fancy paying $200 per month for the privilege. While I can understand videos being watermarked for free users, I don’t think that it’s fair to pay $20/mo and still not be able to remove this.
Many platforms now show when something is AI-generated, so the watermark isn’t adding much extra credibility. I understand it’s brand marketing, but I could easily crop the free watermark with free online video trimmers. It’s also quite easy to resize content these days, and it’s not much more time-consuming than exporting without a watermark.
4
Misreading Prompts
Unless I used the Storyboard feature (which admittedly was excellent), Sora often misread my prompts. I needed to make at least three edits each time I wanted to create a clip. Considering how long it takes to generate videos, you can see how this quickly became a problem for longer-form content creation.
I know the software will improve as it receives more information, but I’m still finding this frustrating.
5
Difficult to Create Still Shots
I’ve tried many different camera angles with Sora, and it’s quite good at creating shots where the camera is moving. Sometimes, though, I want the camera to stay in one place and focus on the main subject. In this area, Sora doesn’t perform very well.
It still seems like Sora is in the “beginner filmmaker” phase, where it thinks that anything that looks cinematic is automatically good. The software regularly ignores my requests for the camera to remain still, so I gave up trying.
Even when feeding Sora a video with the same camera angle I wanted, it didn’t seem interested.
6
No 1080p Videos Without a Pro Subscription
Unless I get ChatGPT Pro, I’m capped at 720p videos. And while many people wouldn’t notice the difference on a computer or phone screen, I still find this limitation quite frustrating.
I could see ChatGPT Pro being worth the money for professional creators who want to make ads. However, I’m a casual user who has no interest in using Sora professionally. Someone could argue that I therefore don’t need 1080p, but I’d still like the option.
Even without access to 1080p downloads, implementing a couple of useful Sora tips will drastically improve their quality.
7
Login Issues
When I first signed up for Sora, I used the same email associated with my ChatGPT account. However, the app didn’t initially notice that I already have a Plus subscription. So, I needed to close the browser and try again.
Login issues are among the most common problems I’ve encountered with new software and websites in all niches. While I expect OpenAI to fix this issue soon, it is still worth considering if you’re signing up. Don’t double-charge yourself.
8
Sora Struggles With Basic Actions
Sora is good at creating videos without object interactions. For example, it’s very proficient at making landscape-style shots, though I can still tell that these are AI-generated.
Unfortunately, Sora feels limited when you need people to do things. For example, opening doors and eating food often look unnatural. I know that AI will probably never be perfect in this respect, but such errors being less obvious would be nice.
However, I recommend addressing this issue where possible. For example, a specific AI prompting formula can yield better results.
While Sora will inevitably improve over time, the early version has some big teething problems to fix. Issues with prompting can dramatically change video production quality, and annoyingly, I need to wait even longer to fix these.
