Today I tried two more video agents, both of which recently raised a lot of money and use the same prompts as Seko. Medeo produced results in just over ten minutes, incredibly fast, and it even supports voice editing. I modified it twice and solved the issues of character consistency and voiceover. It's so intelligent! On the other hand, Flova took several hours to run and ultimately failed to produce results... I seriously doubt whether this can really be called an agent. The difference is just too big.
Seko's version is also included as a reference. Some details were lost during editing, but overall it was completed successfully, and it's very stable and fast.