Sora definitely has its moments of brilliance, but it tends to forget its purpose at times. After spending considerable time using it, I can identify some significant flaws.
1
No Long-Form Videos
One of my main complaints about Sora is the restricted video length, limited to just 5-20 seconds. This duration barely qualifies as short-form content or B-roll. It’s challenging to unleash creativity in such a short timeframe, making it primarily useful for testing ideas rather than creating polished work.
As a ChatGPT Plus user, I’m capped at 10-second videos. I can’t justify the $200 monthly fee for the Pro version just to access 20-second videos. If it were the absolute best available, I might see the worth, but in its current state, I find it lacking in value.
2
The “Subtle” Remix Feature Does Very Little
There’s a big difference between subtle editing and no editing at all. Unfortunately, Sora’s dedicated subtle remix feature seems to fall into the latter category.
Every time I used the subtle remix option, I entertained switching to a different platform. I didn’t notice any changes in the edited videos I uploaded; if anything, the image quality suffered.
I hope OpenAI resolves this issue as Sora continues to develop. However, the current version of the feature is barely functional.
3
Watermarked Videos
While I know I could remove the watermark by upgrading to ChatGPT Pro, I’m not keen on shelling out $200 a month for that option. I understand why watermarking exists for free users, but it doesn’t seem right that I still can’t remove it after paying $20 a month.
In today’s landscape, many platforms indicate when content is AI-generated, so the watermark doesn’t add significant credibility. I recognize it’s a marketing strategy, but it’s simple to crop or resize content using free online tools. It’s no more time-consuming than exporting without a watermark.
4
Misreading Prompts
Unless I utilized the Storyboard feature, which was pretty impressive, Sora often misinterpreted my prompts. I usually ended up making at least three adjustments for each clip I wanted. Given the time it takes to generate videos, this can quickly become a hassle for longer content.
I’m aware the software is likely to improve with more data, but I’m still finding this very frustrating.
5
Difficulty in Capturing Still Shots
I’ve experimented with various camera angles using Sora, and it does well with moving shots. However, there are times when I want the camera to stay fixed on the subject, and that’s where Sora struggles.
It seems like Sora is in the early stages of development, treating everything with a cinematic look as high quality. Unfortunately, it frequently ignores my requests for stationary shots, which has led me to abandon that approach entirely.
Even when I input a video that matched the desired angle, Sora still failed to cooperate.
6
No 1080p Videos Without a Pro Subscription
Without ChatGPT Pro, I’m limited to creating videos in 720p. While many might not notice the difference on smaller screens, this limitation is still quite annoying for me.
ChatGPT Pro might be valuable for professional creators looking to produce advertisements, but as a casual user with no interest in using Sora for commercial purposes, I’d still appreciate having 1080p as an option.
Even without access to 1080p downloads, implementing a few useful tips for Sora can significantly enhance video quality.
7
Login Issues
Upon signing up for Sora, I used the same email associated with my ChatGPT account, but the app didn’t initially recognize my existing Plus subscription. I had to close the browser and try again.
Login issues are a common nuisance I encounter when dealing with new software and websites across various fields. While I anticipate OpenAI will resolve this issue soon, it’s something to keep in mind when creating an account to avoid being charged twice.
8
Sora Struggles with Basic Actions
Sora excels in creating videos without any object interactions. For instance, it’s quite adept at generating landscape shots, but the AI nature is evident.
However, Sora seems limited when it comes to depicting people performing actions. Simple tasks, like opening doors or eating, often appear awkward. While I understand AI may never achieve perfection, it would be nice to see fewer blatant errors.
However, I suggest taking measures to improve this aspect where possible. For example, using a specific prompting technique can yield better outcomes.
While Sora will inevitably evolve as time goes on, its early version has a number of significant issues that still need addressing. The challenges with prompting can dramatically impact video quality, and it’s frustrating to wait longer for improved options.
