Constantly getting failed generations with no explanation why it cannot display my video. Often for the same exact prompt that had no issues generating.

I’m building a super cinematic superhero tale. Action, suspense. All very much like low PG-rated...Honestly, my 6 year old could watch this. I am constantly receiving this monkey graphic with a shrug and an indecipherable message that it couldn’t load my generation and tells me to edit the prompt and try again. But it gives me no direction or information so I have no clue what about the prompt is generating the area. I’ve been doing this a while, so I’m pretty meticulous about my prompts. I color within the lines best I know how. I am asking for nothing offensive or unseemly.

It appears that scenes involving my unconscious FEMALE superhero is incredibly difficult to generate. She’s injured and unconscious and a man is attempting to rescue her. But the mere mention of her being unconscious hinders the process. So, when the action involves the man touching her—picking her up, moving her, carrying her, holding her… as you can imagine, it’s more difficult than pulling teeth without anesthesia. But, sometimes it has nothing to do with her at all. Merely asking for a simple look from my hero, or even just scenic shots might still result in getting this monkey. But, to reiterate, I’m given no information, no direction, no clues as to what has gone wrong and what to fix, so I’m literally working off conjecture here. 

I’m keeping the scenes as simple as possible, trying not to overload each generation with too much action for the model to process. I say all that to say, with all my careful intention, I simply can’t gauge what it is I could be doing wrong or how to fix it. The system should provide more detail than just “try again”. But, also, that monkey has popped up so frequently and even when I attempted to generate a second take of the same prompt. The first generation was fine. But the second time—monkey. I ran it a second time. The generation succeeded. Feels like a glitch.

Please help.

———

Example of failed generation:

The African American man cradles the woman securely in both arms as she sleeps (I have to tell the system that she’s sleeping so it doesn’t block the generation for using the work “unconscious”.). He pauses to catch his breath. His golden eyes suddenly shift and lock onto something just beyond the camera, slightly below eye level. Concern slowly overtakes his expression as he realizes something is approaching. He does not move—only his eyes and head subtly track the unseen threat, and then notices another coming from the opposite direction. Blue veins pulse faintly through the dead trees and fractured ground around them. Cinematic realism. Natural movement. Slow, tense movement. Dramatic, suspenseful tone.

I successfully generated this scene with the same Start Frame with a slight tweak to the action. I simplified this particular prompt and was able to get a successful clip. But, it’s all guesswork. The tweak was so minor, I’m not really sure if I landed on whatever wordage the system needed to process it or if it was just a glitch.

I’m not against safeguards with AI at all. In fact, I’m in full support. But honest creators do need an avenue in which we can successfully operate and build our visions without having to employ forensic-level prompting.

Thank you for your time.