Firefly massively overanimating and not respecting my explicit instructions

[Ru Howe, Stornaway.io]

I uploaded a still of a kayak on water to turn into a video.
I have several times, in increasing exasperation, told Firefly to ONLY make the water ripple slightly and NOT to move the kayak, the kayaker, the man, or anything else in the frame. It started out madly animating the camera, and when I told it to keep the frame static it won't stop adding mad animations to the kayaker, inserting a man's arms in the front of the frame paddling - every time, I tell it DON'T MOVE THE KAYAK, DON'T MOVE THE MAN. DON'T ADD A MAN BEHIND THE CAMERA PADDLING. But it won't listen. I'm 8 prompts in and writing prompts now like 
GO BACK TO THE ORIGINAL IMAGE. DO NOT MOVE THE KAYAKS. DO NOT MOVE THE KAYAKS. DO NOT ADD PADDLING. THE KAYAK MUST BE STATIC. THE MAN MUST BE STATIC. THERE MUST BE NO MOVEMENT IN THE FRONT OF THE FRAME. DO NOT ADD ANOTHER MOVEMENT.
And then like it's trolling me, it makes the man paddle backwards past the camera. 
And then in the next one still makes the man paddle but has put a big STOP road sign in front of the camera. LOL, except it's not funny. I've lost 30 minutes to this and it's so frustrating - such a waste of time and money, paying to train it but it won't be trained. I just wanted a simple easy cinematograph style water and light effect - it's done all that, and then gone rogue and won't listen.

[Nite-Stian]

Hey! This seems really frustrating!!

I do find that Firefly do not respond well to verbs, telling it what to do works worse than describing what you want. Keep it short and specific and if Firefly seems stuck on certain elements try changinc the seed (by manually typing or opening another window). 

Trying this out I didn't have an image of a man in a kayak so I decided to generate one. I see that Firefly seems to prefer adding a kayak to the foreground so this issue might be related to Firefly's training material. Using this prompt in the video generation I was able to come close to what you seem to achieve: "a man rowing slowly in a kayak on a mountain lake" (I accidently chose handheld camera instead of static).

Also please, whenever you get a result you're not happy with, make sure you hit the thumbs down button! This way the team can work on improving the beta model.