As we’ve recently covered, Kling AI is off to a promising start, with the powerful new 2.0 model doing a great job of generating image-to-video projects and working with text prompts to create some funny, unsettling, and downright bizarre creations.
After Kling reached out to hand me the keys to the kingdom, I wanted to test its creativity out with something that’s less a direct prompt, and more a request for something a tad more ethereal.
To do so, I asked it to generate a five-second music video for the song ‘Rain’, by one of my favorite bands, Sleep Token. I wanted to see how far I could push the model in its recognition of the underlying themes of the song, or whether it would default to, well, a lot of rain.
Five seconds is available for free users, and set me back 100 credits per output while using the Kling 2.0 Master model, and I was impressed at just how well it understood not just my prompt, but my negative prompt, too.
Here’s how it unfolded.
The Prompt
The band is known for its dark, almost fantasy-like lyrics, and Rain follows that theme just as much.
With all that in mind, my prompt was:
“Create a sci-fi tone piece for the song Rain by the band Sleep Token, taking care to include rainfall in a cyberpunk-esque city.”
For negative prompts, I was cautious of having too much color. Cyberpunk tends to evoke a certain palette of pinks and purples, but that felt a little too easy. Instead, I wanted something just a little more grounded and brutalist.
“Do not add too much vibrancy to the project, it should feel relatively grey in terms of colour” was my negative prompt.
Kling’s site gave me a rough estimated waiting time of 6 minutes, so I grabbed a cup of coffee and returned to some pretty staggering results.
The Result
I was offered four outputs by Kling, and while this may sound like damning the model with faint praise, I was impressed by the consistency of results.
All four feature a futuristic city and plenty of rain (as was to be expected), and they all do a relatively good job of keeping lighting away from being too bright in any one spot.
I also notice that the car on Output 2 looks almost like it’s in front of a van, before it becomes clear that they’re the same shape.
I was consistently impressed by the way the umbrella in the videos that feature one is transparent, and you can see buildings beyond it in each shot, too.
Because Kling only produces short videos, we are currently unable to make an AI generator create full-length music videos. However, it does seem like that is the direction we are heading in.
These short clips show how far AI video has come already, especially in understanding context. When Sora, the first major AI video generator, launched, it was messy and a bit all over the place.
This shows just how far we’ve come in this area, but what do you think? Any projects you’re considering trying with Kling? Let us know in the comments!
How you can use Kling too
Kling is rapidly growing in popularity, thanks to its impressive text-to-video generation ability.
If you want to get involved and try it out for yourself, check out our guide on how to use Kling.
Not just for video, Kling can also be used to generate images and even sounds. When it comes to video generation, the tool can generate via a prompt, an existing image or by editing a video you already have.
This gives an existing variety in how you can interact with Kling. Try it out yourself and see what you can produce!
