How long is this nonsense concerning not being able upload an image containing ‘people’ in conjunction with a text prompt going to continue?
Whilst I can appreciate concerns surrounding the power of this new technology and potential abuse, OpenAI’s policies are far too restrictive and need to be balanced with the First Amendment over so called ‘community standards’. There are already long standing libel laws in place to deal with potential bad actors so let the rest of us explore creativity in freedom please.
The most bizarre aspect of all is that Sora will not even accept an image made by Dall E3 which is obviously fictional and made by its sister company for goodness sake! If it’s so clever, why doesn’t it know that?
The concept of video AI is fascinating and in all likelihood will be the closest we will ever come to inventing a time machine! Or, to paraphrase Sir William Lyons, ‘the closest we will come to creating something that is alive’.
I signed up for GPT plus purely to get Sora. If it can produce, I would upgrade to Pro. So far, I don’t see evidence that it can and I am also a paying subscriber to the two other competing products that beat it to market and despite their own growing pains are producing stellar results by comparison. I shall have to take a hard look at keeping GPT when the subscription comes around for renewal if things don’t improve.
You need a pro subscription to use images with people.
Why? Well. Should be obvious
I was in the same boat. Unfortunately not being able to test it with people images, along with some other issues left me deciding to not even bother with pro.
That’s strange. I recall in the Sora interface it explicitly mentioned that only pro users could use uploaded images of people. Can’t find any information for it anywhere besides some articles of OpenAI claiming that a “subset” of people are allowed to do it.
I also have seen my own image not work, but then when passed to someone with a pro subscription it worked (well, worked as in generated…)
Hi guys, I have a pro subscription. I’m experiencing with Sora since yesterday. Each time I upload an image with people, it doesn’t work most of the time. The results are still or f*cked up, it doesn’t keep the consistency of the input but imagine things in a weird digital 3D render instead and the scenes have no sense even if I add a text prompt to the image input. I’ve tried everything, I generated more than 30 videos with images inputs and none is even slightly good. I’m so disappointed and a bit angry actually, it’s terrible to make people pay this much for these kind of results, I really hope they will do something to improve this and really quick. If someone has any tips, don’t hesitate, I’m desperate :')
I’m sorry for you but I feel relieved to know that I’m not alone, I thought I was doing something wrong and was going crazy honestly. I even struggle to get good results with simple texte prompt. I can’t get my head around on how to get a certain aesthetic for the render. I get a 3D shitty style every time and the movements are totally random if it last longer than 5s (even with clear instructions). I tried adding custom presets but no matter how I write it, it doesn’t work. I got only one good result and it was using the “film noir” preset.
Anyway, I’m going to keep trying and searching but it feels like they released it wayyy to soon
Same here; I’m a professional graphic designer. One of my specialties is ‘Talking Avatars’. A photo-to-video tool, refusing to create a video ‘because the photo contains people’ is honestly the strangest thing I have ever experienced, especially when at signing up I have declared to have ownership of all uploaded material. This is over-the-top restriction. Pretty convinced a lot of users will sign-off on this kind of controll.
I had hoped to use Midjourney images I had created based upon my own artwork to use in a creative way with Sora generated videos, but there are people in those images (that was kind of the point). There’s no way I would consider a Pro subscription unless I know what I want to do is possible. From what I hear, it sounds like it’s not, really. Disappointing. I had high hopes for Sora.
This is disappointing to hear. I was hoping to use Sora primarily for generating specific stock footage that accurately blends with footage I’ve shot myself when needed. Seems I won’t be able to do that if the footage I’m attempting to blend with contains actors.
I mean, its hilarious that I can’t use my A.i images with people in them, as its being flagged for having people.
20 bucks wasted. How long before you can get a refund? There is 0% of a pro license if I can’t test the damn thing out, but apparently it happens to Pro’s too?
I’m just catching up to this, I’d just cancelled my Pro subscription and had two days left on it when they released Sora.
All the alpha testers dropping super cool ad creatives featuring hi-res close-ups of people convinced (and inspired) me to renew.
I’m finally getting time to play with it, only to learn that the users who posted all those animated images with people in them are part of a secret squirrel club that I’m not invited to.
They need to lead with that nonsense, not “Oh by the way” it when the user goes to upload an image.
Business is business, I get it. But this is sleazy and becoming typical of them, and at this point I don’t know why I expected any different. Time to cancel for good this time.
I spent some time doing it in runway and had some decent results but was only 720P. Sora has been a huge disappointment. I wonder who gets the great features that were featured in their promotional materials.
I’m in a similar situation to you, but until now, my subscription has expired. I haven’t used Sora even once.
Other similar products are not worse than Sora. Regarding consistency, coherence, and semantic understanding, Sora was not there for nearly one year.
Kling, Hailuo, and other models have given people a lot of surprises.
If Sora does not update and improve, it may be gradually abandoned and become a doomed outcome.
my take
I am NOT paying $200 for Pro subscription, especially after learning it doesn’t even work well
I tried runway, it’s ok, but it doesn’t feel real
my personal favorite is Kling AI, it is very real, the best I have seems so far among other image to video AI. of course, having bugs at times is unavoidable
granted, it is from China, some people may have problem with it because of this. but otherwise, I think it has a very good AI engine/algorithm
I can confirm that don’t work on any image, for example, I’m a 3d artist and I tried to animate some idle animation on my characters ,illustrations etc…, but it does not work, and not just on humans but also on anything else, it says it violates some IP rights (how funny) Its practically useless
Doesn’t work with my pro subscription either. And I’m using like 2D concept art images and renders of a 3D model lol Kinda frustrating as you’d think they’ve had enough of my money to know I won’t misuse this tool
Even when it accepts a pic of a person, only a tiny rectangle in the middle will be used, so if pic is head to toe, the rectangle only covers the torso and you cannot even move the rectangle up or down.
Not worth going pro subscription for it at the moment.
but I feel relieved to know that I’m not alone, I thought I was doing something wrong and was going crazy honestly. I even struggle to get good results with simple texte prompt. I can’t get my head around on how to get a certain aesthetic for the render. I get a 3D shitty style every time and the movements are totally random if it last longer than 5s (even with clear instructions). I tried adding custom presets but no matter how I write it, it doesn’t work. I got only one good result and it was using the “film noir” preset.