At the beginning of 2024, OpenAI showed the world the capabilities of its new neural network Sora. She knows how to create videos based on text descriptions, and does it at a professional level – looking at examples of Sora’s work, you might think that these are stills from movies and cartoons. We have already played enough with the capabilities of neural networks for generating text and pictures, now we all want to create videos. But access to Sora is closed, and no one knows for sure when the neural network will become available to everyone. Fortunately, other companies are not so greedy and will allow you to generate videos using their services. If you do everything correctly, the quality of the videos is at the level of Sora. Are you ready to create your first video using Neuron? Roll up your sleeves and repeat after us – now we’ll teach you everything.
Still from a video created by the Dream Machine neural network from Luma AI
Free neural network for video creation
The best analogue of Sora is the Dream Machine neural network, developed by Luma AI. It was introduced to the world in mid-June 2024 and is free, but with restrictions. Free users can generate 30 videos per month, no longer than 5 minutes. To increase your limit, you can buy a premium subscription to Luma AI Dream Machine, costing from $30 to $500.
Prices for the premium version of Luma AI. Source: habr.com
Neural networks have been able to generate images for a long time; they do it quickly and in masterpiece quality. Creating multiple interconnected frames has always been difficult for artificial intelligence. The neural network from Luma AI was able to surprise us – the videos are such that some people may mistake them for the work of professional operators and artists.
Read also: The Sora neural network turns text into video – you can Should I use it right now?
How a neural network makes video from text
The Dream Machine service from Luma AI is available to everyone, even residents of Russia. To use it, you don’t need to know much – everything is done in a couple of clicks.
To generate a video using a neural network, you need to go to the Luma AI website and click on the buttonTry Now. In the next window, the service will ask you to log in to your Google account; you must do this, otherwise the system will not allow you to continue.
Dream Machine main page, the “Try Now” button is located at the top right
After logging in, the fun begins. There is an input field in the middle of the screen: there you need to write what should happen in the video. You need to write in English (!), so it’s better to use a translator.
Dream Machine Prompt Input Field
As an example, let's try to create a video with an astronaut sitting on a bus. The description should be as detailed as possible and have a style description, so we'll try the prompt: an astronaut sits in the bus and takes off his spacesuit, sunlight falls on him from the left, it's summer outside the window, photorealistic, hollywood movie.
Before generating, you need to check the box in the “Enhance prompt” item, which significantly improves your description in the background.
As a result, we really got a video with an astronaut on a bus, which looks like an excerpt from a Hollywood movie. Only we asked him to take off his spacesuit, but this did not happen, it was necessary to describe the movements in more detail. Look at the result.
Astronaut riding on a bus
As a second example, we tried to create a video of a 25-year-old girl with big green eyes in a post-apocalyptic desert. She is clearly younger, but the video really looks like an excerpt from the movie “Mad Max” or the acclaimed TV series “Fallout.”
Kopeck piece with a desert background
Read also: 5 neural networks for creating pictures that few people know about
How to live memes using a neural network
The Dream Machine neural network from Luma AI can also turn photos into videos. To do this, next to the input field you need to click on the photo icon and upload the file. After this, you need to write a description as in the instructions above and wait for the result.
Button to upload a picture
For example, let's try to animate a picture from an article in which we learned how to use the Bing neural network from Microsoft. It shows a girl in a white dress with an umbrella in her hand – let's make it rain, blow, and make people pass by behind. The prompt looks like this: a girl with an umbrella in her hand stands in the pouring rain, drops fall from the umbrella, people pass by in the background, dramatic film.
As a result, we got a great video, but for some reason the neural network did not draw rain. Check out the result.
Girl with an umbrella in the rain
Many people use the Luma AI neural network to revive memes. Let's try to turn a meme with a flying cat into a video. Upload a picture and write a description: a ginger cat flies in the air, its fur wriggles in the wind, the background flies back, epic shot.
Video with a flying cat
Once again, the result was not quite what we wanted, but the result is still impressive. Neural networks are developing very quickly, and after six months or a year, every person will be able to create masterpieces.
As you can understand, the neural network for creating videos from Luma AI does not always understand what is required of it. In addition, you can notice distortions in the video. Well, the main disadvantage is that there are restrictions on the number of videos generated per day.
Clip created using a neural network
The presence of neural networks such as Dream Machine from Luma AI increases the chances that that OpenAI's Sora will soon be open to everyone. If competitors make their neurons free, the ChatGPT developers will also have to abandon plans to make Sora paid – such rumors have been circulating since the presentation.
Did you manage to make a video using the Dream Machine neural network from Luma AI? Write in our Telegram chat!
If you combine all the videos created by the neural network, you can get an excellent clip. Music for it can be written using artificial intelligence, read our article “Neural networks for creating music: free services available to everyone.”