Comparing AI animation settings

In 2024 we spent some time for research and investigation on the workflow with two Midjourney generated futuristic architecture renderings in a benchmark test which takes a focus on comparing animation settings inside Krea.ai, with analysis of render times, fps and size settings by using the render-engines Hailuo, Luma, Kling (Standard), Kling (Pro) and Krea Animator and comparing the internal results with an external counterpart of using RunwayML. 

Recheck every 3 months

The rapid development of new AI-Video tools strongly forces us to compare and monitor the existing and upcoming applications at least every 3 months. AI agents can help to identify and compare these apps, even if the human proof of comparison is essential to avoid wrong results.  

Hyper, hyper - wait you are still the most reliable tester

There is a simple hint how companies try to gain attention, especially in terms of videoproduction this is interesting. E.g. selling a framerate of 60fps sounds already great, but after loading the animated result into a meta-analysis, it turns out that the real fps is only 47fps. With a jump down of 13 fps, this is an important topic for producing smooth running scenes that still can be pitched down. 

But even if you trick the user into thinking that 10 seconds are being generated and the result is 10:47 seconds, this makes time-based editing more difficult, so the hype around new inventions is often less reliable because everything happens too fast. Developers should keep reliability in mind before succumbing to time pressure due to strong competition. So checking the results yourself is still the most reliable solution.