I started IA to kinda chart the progress of AI, and my ability to stump “them” on any given topic. While I could do it initially (see my first post from 2023), it’s impossible now so I’m just kind of dabbling in AI things that catch my attention when I have a minute.
Generative AI Video
I’ve been doing a ton of stuff with generative AI video, not the least of which is BibleBitz.com, a cinematic experiment where passages of Biblical scripture are presented in new and creative ways that just wasn’t practical just a couple of years ago.
While there are MANY AI video platforms cropping up almost daily, I’ve been using KlingAI and RunwayML almost exclusively because they were the most accessible, and affordable while folks like Sora were (and are still IMO) play catch up. Some of the outputs from Sora I’ve seen are amazing, but so far, I’ve experienced loading problems and general unreliability issues that I don’t want to deal with.
Runway Gen 4 vs Kling Master 2.1 Generative AI Video
Kling Master 2.1 vs Runway Gen 4
I started IA to kinda chart the progress of AI, and my ability to stump “them” on any given topic. While I could do it initially (see my first post from 2023), it’s impossible now so I’m just kind of dabbling in AI things that catch my attention when I have a minute.
Generative AI Video
I’ve been doing a ton of stuff with generative AI video, not the least of which is BibleBitz.com, a cinematic experiment where passages of Biblical scripture are presented in new and creative ways that just wasn’t practical just a couple of years ago.
While there are MANY AI video platforms cropping up almost daily, I’ve been using KlingAI and RunwayML almost exclusively because they were the most accessible, and affordable while folks like Sora were (and are still IMO) play catch up. Some of the outputs from Sora I’ve seen are amazing, but so far, I’ve experienced loading problems and general unreliability issues that I don’t want to deal with.
Prompts for Generative AI Video
My own personal approach to video prompts doesn’t always yield the ideal results, but the other side of that coin is that unexpected results can spark creativity and be a great source of inspiration, even if it isn’t germane to the task at hand.
I’ve created a project in Anthropic’s Claude I call “Prompt Central”, where I’ve uploaded to the PK the prompt best practices specific to each platform. In fact, I even have Midjourney in there as well, which I’ve typically used for stills, but will have to update in the future as they recently added video. I’ll have to work Midjourney into the equation on the next comparison 😉
RunwayML Gen 4
The prompt:
“First-person perspective racing forward along the desert path at high speed. Camera hurtles between scattered rocks and ancient stone pillars. Sand particles kick up and swirl in the wake.”
The result:
KlingAI Master Mode 2.1
The prompt:
“First-person perspective camera rockets forward along the winding path at high speed. Sand particles whirl and kick up from the ground as the camera races between scattered rocks. The ancient stone pillars rush past on either side. Motion blur streaks across the desert landscape as the camera maintains breakneck speed through the canyon valley.”
The result:
Conclusion:
As is typical of most of the AI videos clips I generate, I think KlingAI is the clear victor. From the standpoint of aesthetics, they both look pretty good, but RunwayML seemed to chose its own path of least resistance by creating new path(s) convenient to the camera movement. KlingAI took the initial path and created a single path that best presented as the logical, and organic progression.
I’m going to be doing more comparisons as the opportunities arise. One test is far from conclusive, and both platforms are constantly improving. There are some applications where RunwayML excels and wins out over my KlingAI option, especially when depictions of English words are concerned. That said, at the time of this post, I’d say that in the day-to-day, real world use of generative AI for the purposes of my own projects, KlingAI consistently yields better results when pitted against RunwayML. More on this and others, later 🫡