AI can make videos now. The video below was made with OpenAI's "Sora."
-
Welcome to the Fantasy Writing Forums. Register Now to join us!
AI Trends
- Thread starter Demesnedenoir
- Start date
Fyri
Inkling
I mean... I just looks like someone copied an already existing video, maybe added another dog from a different video and pasted the gifs so to say onto a picture of a mountaintop with some added pictures of a blanket and mic.
Fyri
Inkling
This one is interesting though. I just hope that whatever happens, we find ways to make our new abilities also ethical, at least by law. Idk.
Yeah it "just" created two dogs, a background and added lifelike movement in what likely amounted to a couple of seconds at most. You're saying that as if it's some ordinary thing. The speed at which people get used to programs producing something that would have taken an hour or so and required real footage is madness.
Last edited:
Finchbearer
Istar
Have you not seen all the film trailer AI videos? Like lords of the rings but Wes Anderson, or Lord of the rings, but in Berlin…
There's a difference between taking AI images and adding facial movement to it (which most of those videos did) and generating a video from scratch based on a single prompt.
Finchbearer
Istar
So it’s a pioneer video from a single prompt?
Yes, these (including the video linked by Fyri) are based on one prompt each. A single sentence went into making them, which is pretty revolutionary.
Prince of Spires
Auror
If you're doing TikTok videos or Facebook ads, then this can actually be a massive development. Video ads were just too expensive to be worth the effort for allmost all indie authors. Being able to create something like this for your book with an hour or so of work (it takes some playing around with the prompt probably to get some decent results) is massive. For an ad or a Tiktok video it doesn't need to be perfect. It just needs to be good enough to get you noticed by people. And these certainly are that.
The examples above showcase what a first output might look like, but some curated examples can be found on the dedicated Sora page on OpenAI's website: Sora: Creating video from text
Some that I found particularly impressive:
Some that I found particularly impressive:
- Thread starter
- #131
Demesnedenoir
Myth Weaver
A couple of those are impressive as hell.
- Thread starter
- #132
Demesnedenoir
Myth Weaver
Adobe Photoshop's Generative Fill is "AI" similar to Content Aware Fill, but on steroids, is a game changer when working with your covers and translating artwork from ebook, to paperback, to hardcover. I had an image end up on the spine in a bad location, so I just split the cover, dragged a piece to the left, selected the open space, and hit Generate... Boom, the space is filled, and the face getting covered by text is now safely on the back cover. And I could move it to the front cover if I wanted to. Crazy simple.
Another area with some interesting potential is in teaching. I could show--now either static or as video--an engaging (note I did not say realistic) picture of a medieval village. It would be miles ahead of the classic line drawing diagram of a village, and would put to shame those painstaking reconstructions of things like Amiens Cathedral.
I'm waiiting for Sora to go more public, when I'll ask it to show me a medieval village. Then I'll ask it to show me an 8th century village, then a 13thc village. Then I'll ask it for a Scottish village, a Sicilian village, a Polish village. Even more fun, I'll ask for an 11thc Pomeranian town. Or a Wendish village and see how far it gets with that.
I opened by saying there's potential. That is a neutral adjective, and there's potential to mislead or to perpetuate misconceptions as much as there is to "educate". The way technology affects pedagogy continues to fascinate me.
I'm waiiting for Sora to go more public, when I'll ask it to show me a medieval village. Then I'll ask it to show me an 8th century village, then a 13thc village. Then I'll ask it for a Scottish village, a Sicilian village, a Polish village. Even more fun, I'll ask for an 11thc Pomeranian town. Or a Wendish village and see how far it gets with that.
I opened by saying there's potential. That is a neutral adjective, and there's potential to mislead or to perpetuate misconceptions as much as there is to "educate". The way technology affects pedagogy continues to fascinate me.
CupofJoe
Myth Weaver
One thing I've found out working with AI, is that you have to know what the answer should sort of look like before you can ask the right questions to get you the answers you want. And that is assuming the information is there for it to work with.
- Thread starter
- #135
Demesnedenoir
Myth Weaver
If/When it gets to the point of historical accuracy I'll lose a lot of hours playing around with that. Of course, I might be dead by the time that happens and have plenty of time to waste.
I appreciate your unceasing capacity for positivity Skip. I'd rather have real human beings be compensated for the creation of such documentary-style films, but if individual teachers can utilize this AI in class, that's at least one tiny light in a gurgling tar pit of impending propaganda, revenge films, false evidence, mass layoffs, broken dreams and more that Sora will create once it hits the market. I mean that genuinely. I was struggling to find even one bright side of this tech, and now there's at least something.
- Thread starter
- #137
Demesnedenoir
Myth Weaver
"AI" in use in Illustrator... I'm working on a quote for new countertops and downloaded Illustrator to use as a quickie CAD, and lo and behold! AI.
I didn't say historical accuracy because almost any illustration of anything medieval isn't completely accurate. Indeed, we could have a whole deconstructionist conversation about what constitutes historical accuracy. But I don't what that!
As a teacher, I thought of images as tools for teaching. For example, take any map of medieval Europe. It will always show borders. That's not "historically accurate" and in some ways it's misleading. So I tried to make time to talk about what constituted a border in pre-modern times, and what a "border" might mean when the term wasn't King of France but rather King of the Franks.
That just a tiny example. Next month I'm giving a talk on the conversion of pagans in 12thc eastern Europe. We have descriptions of the baptism process and we even have some really poor illustrations. Now, is it wrong to have AI create a picture? It's unrealistic to expect a human to do so, as this is merely a one-time presentation to a small crowd of oldsters. I could take the poor illustration, which is too small for decent projection anyway (which at this point is probably what I'll do, accompanied by energetic hand waving and pantomime), or I could have AI draw me something that is at least as historically accurate as those medieval maps, and would like draw more people into the conversation, not to mention make the whole presentation more memorable. And what good is teaching that goes unremembered?
I can run in lots of directions with this. There's only one contemporary picture of Emperor Frederick II, and it's almost cartoonish. Why not take composites from some of those reconstructed medieval faces and create a more vivid (and more memorable) picture of him? Before anyone objects--which I can do because I've got the keyboard just now--I will point out that there are any number of illustrations and engravings from the 19thc and 20thc of this and other monarchs. All of them more or less fanciful. What we were wont to call the "artist's conception". Uncounted history books have been adorned by such images, though we're rather more conscientious about it all these days than we once were. Still, I can picture (pardon) a time when AI provides a fair amount of the illustrations of the typical history textbook, or even of a scholarly monograph.
As a medievalist, I'd be delighted if the entire corpus of images from those thousand years were fed into AI. Better that than the racket being perpetrated by Getty et alia, who have taken all these clearly-out-of-copyright pictures, watermarked them, and will release them only for cash on the barrelhead. Instead, let AI feed off our common past and construct for us images as each teacher or researcher finds need. Free of charge, you jackals pretending to be custodians of art.
Excuse me. I seem to have got some rant on my shirt.
As a teacher, I thought of images as tools for teaching. For example, take any map of medieval Europe. It will always show borders. That's not "historically accurate" and in some ways it's misleading. So I tried to make time to talk about what constituted a border in pre-modern times, and what a "border" might mean when the term wasn't King of France but rather King of the Franks.
That just a tiny example. Next month I'm giving a talk on the conversion of pagans in 12thc eastern Europe. We have descriptions of the baptism process and we even have some really poor illustrations. Now, is it wrong to have AI create a picture? It's unrealistic to expect a human to do so, as this is merely a one-time presentation to a small crowd of oldsters. I could take the poor illustration, which is too small for decent projection anyway (which at this point is probably what I'll do, accompanied by energetic hand waving and pantomime), or I could have AI draw me something that is at least as historically accurate as those medieval maps, and would like draw more people into the conversation, not to mention make the whole presentation more memorable. And what good is teaching that goes unremembered?
I can run in lots of directions with this. There's only one contemporary picture of Emperor Frederick II, and it's almost cartoonish. Why not take composites from some of those reconstructed medieval faces and create a more vivid (and more memorable) picture of him? Before anyone objects--which I can do because I've got the keyboard just now--I will point out that there are any number of illustrations and engravings from the 19thc and 20thc of this and other monarchs. All of them more or less fanciful. What we were wont to call the "artist's conception". Uncounted history books have been adorned by such images, though we're rather more conscientious about it all these days than we once were. Still, I can picture (pardon) a time when AI provides a fair amount of the illustrations of the typical history textbook, or even of a scholarly monograph.
As a medievalist, I'd be delighted if the entire corpus of images from those thousand years were fed into AI. Better that than the racket being perpetrated by Getty et alia, who have taken all these clearly-out-of-copyright pictures, watermarked them, and will release them only for cash on the barrelhead. Instead, let AI feed off our common past and construct for us images as each teacher or researcher finds need. Free of charge, you jackals pretending to be custodians of art.
Excuse me. I seem to have got some rant on my shirt.
pmmg
Myth Weaver
How can AI have historical accuracy? I doubt we can ever agree on what is accurate.
(Well, Skip beat me to it, I think).
Last edited:
Finchbearer
Istar
Plus there will always be something to be pointed out as historically inaccurate by someone 
As far as AI goes, it just simply isn’t at a point of being reliable in the accuracy corner. You still have to go and double or even triple checks your facts. And even then there is always going to be misinterpreted information out there.
As far as AI goes, it just simply isn’t at a point of being reliable in the accuracy corner. You still have to go and double or even triple checks your facts. And even then there is always going to be misinterpreted information out there.