Okay, so I jumped into Luma Dream Machine this week, having only really used basic WordPress and a bit of GA before. I'm *very* new to all this AI video stuff and honestly expected to fail spectacularly.
My biggest surprise? It actually listened to my first prompt ("a cat astronaut floating in zero-g, cartoon style") and made something... recognizable! It wasn't perfect, but it existed. My biggest frustration is the inconsistency. That same prompt, run again an hour later, gave me a weird three-eyed creature. Also, the 4-second limit feels like it cuts off every idea just as it gets going.
Is this just me being a noob? Are there some secret settings I'm missing to make generations more stable? I'm determined to get something usable for a landing page background, but it's a wild ride so far 😅
Welcome to the generative "art" rodeo. You're not a noob, you've just discovered the core feature: inconsistency masquerading as creativity.
There are no secret settings. The 4-second limit isn't a mystery, it's a hard cost/GPU boundary. Your three-eyed cat proves the model is stochastically sampling a latent space, not following a recipe. That's the game.
For a landing page background? Good luck. You'll burn more credits chasing consistency than you would paying for a stock video. Sometimes the "wrong" tool is just the wrong tool.