Text to Video: The Current Hot Generative AI Topic
I know pretty much nothing about video. I mean, other than recording a basic video with my phone.
But I heard text-to-video is the newest hot thing in generative AIs, and wanted to try out one, just to see what we are dealing with.
I typed "text to video AIs" in Presearch, and took the first non-ad result, which was Fliki.ai.
Made a free account, which gives me access to 5 minutes of credits per month. I'm not sure when these credits will be used because I created a video that should have used some credits. Maybe the credits accounting is updated at the end of every day or something.
Anyway, as I was saying, I created a test video from an idea, as it is called on Fliki. Instead of an idea, one could use alternative options like a blog post, a Powerpoint presentation, a tweet, or a product to start with. Or even create a blank file. But why would you use this tool, if not for the AI?
The idea I fed the AI was very basic: "enjoyable walk on the beach". Three words and 2 fillers. With that, it created a 1 minute+ video, with several scenes, a female voice-over (changeable), and a subtitle throughout the whole video piece.
Background audio of the video, the script, and videos (I chose stock video, but could have been AI-generated art) on every scene could have been changed, and I even changed a few scene videos that didn't exactly match what I had in mind. It used, for example, a video of winter scenery in one of the scenes, probably because two people were walking in the video.
Screenshot from inside Fliki.ai
More scenes can be added, deleted, or their order changed.
So, what was the AI's part? Everything. I just gave it the idea "enjoyable walk on the beach". The audio track, voice-over, script, and video selections for scenes, and subtitles, were all done by the AI, after a few seconds of processing.
Yes, I've done a few tweaks afterward, and that wouldn't pass for a professional video without someone who knows what he's doing operating the software, but it shows the power of this tool (and others like it).
Screenshot from VLC media player on my laptop
Here's something I remarked while checking out its menu. It has something called "voice cloning", available for premium users: 2 minutes of you talking and it can say anything using your own voice. The problem I see here is if someone starts cloning other people's voices and using them to say whatever they want (and I've seen these kinds of incidents start to multiply).
We are at a time in history when generative AIs change at an exponential rate and will have a profound impact on all of us, the economy, and society. Be ready!
Want to check out my collection of posts?
It's a good way to pick what interests you.
Posted Using InLeo Alpha
AI will impact on the economy positively and that’s sure but I think we should never forget that it is gradually taking our jobs
No, we shouldn't forget that. But they are coming. The question is, do we want to know how to use them while we can or be afraid of them?
My goodness, now even video can be generated from text prompt.
Yeah, and from what I've seen it's also used to generate deep fakes. The US elections this year will be fun.
AI impact on our life is going to be massive, and it got started.
That's inevitable. It will be in almost everything.
yes, no doubt about it.
That's good for creating content on YouTube. I will try it out as well. I'm thinking of dedicating some time to creating motivational(and anything related) YouTube videos and this might help.
With the right prompt and some tweaking afterward, the videos can be ok, so yeah, they might turn out pretty good for motivational videos.
So much that Artificial intelligence is really changing in our world and it is of a concern
It will unfortunately probably lead to some troubled times ahead before everything settles down.
Shared on X for more engagement
https://twitter.com/jewellery_all/status/1743727072764149952
Thanks.
I still believe we must be cared about the work Artificial intelligence is really doing in our society presently now
AI is a field that must be treated with care, and I'm pretty sure it's not always the case. It's the first industrial revolution where the tools will outsmart their makers - humans.
With how powerful generative AI has become, it seems a majority of media being AI made is coming sooner that we think. If you think about it, videos are just a bunch of images having minute changes in quick succession. If AI can recreate works of art, and add on it using the same art style, making videos from a single image would be easy.
Text to video might be difficult to use, unless one is able to describe everything perfectly. I think the current best progression would be text to image, then image to video.
I believe the majority of articles in mainstream media are written using AI and only brushed here and there by a human.
One deterrent in this area is the higher processing power needed to create a video rather than text or image. The longer the video, the more complicated it becomes. Even more, the longer the scenes, the higher processing power it is needed, if the AI needs to create the frames coming from an image. I don't see this kind of use case exploding very soon on a large scale due to hardware limitations, where the AI creates scene frames starting from an image.
Another possible deterrent is training the AI to follow a consistent idea over a long series of frames that make up the scene. It's different than generating one image following a particular style. Right now, they are rather rough despite improving the productivity of simple tasks by a lot compared to what it would take a human.
It doesn't have to be perfect, but probably the more precise the better.
After only using this program briefly, my impression is that text-to-video is at a stage where human updates post-generation are critical to making the video's message more consistent overall and avoiding awkwardness. And still, a human-made video from someone who knows what they are doing would probably be better at this point.
That makes sense. As for the processing power limitation, it can just be smaller clips, and then cut together in a video editor. Big corporations wouldn't have this problem as well if they make use of super/quantum computers.
Yeah, but quantum computers with some utility are still far out, and would they use a supercomputer to make it public to the masses, or behind a low-end paywall?
I think big companies will monopolize those. I can see Disney/Pixar shifting towards it in the future, and maybe big anime companies. For the masses, high end GPUs are getting better every year. Connecting multiple units together, like how they used to in mining, can make it viable.
Ah, yes. AI will be a playfield where big tech giants will make the rules.
I think when this feature becomes of age, it will upend much of the movie industry. More creative people will have the ability to 'easily' create interesting long form video content.
The first ones that will be disrupted are the commercials because they are shorter and use stock videos or static images intensively.
Before it would affect the movie industry to a great degree, I see something like TV series (with unrelated episodes) or documentaries being targeted because they are shorter per episode.
I am curious once the movie industry centralization is broken, will we be able to find great movies created by unknowns? Or the feeding and trending systems will remain the same?
Ai sure is changing things quite fast. It would definitely be concerning is people started using other people's voices. I think we are still at the stage where people can figure out if it's AI generated or not but technology is improving fast. So it might become quite hard later on in the future.
We will see how much people can tell the difference now. I'm pretty sure deep fakes will be used intensively in this year's elections in the US.
https://twitter.com/LovingGirlHive/status/1743927519013986370
This technology is becoming very popular all over the world and is very special for students as it tells them many special things that help them a lot in their studies.