Camera animation can not be disabled at all

Hey ​@bacizone 

This is definitely frustrating. We are here to help and provide some insights.

To add to what ​@droopydog500 mentioned
I tend to add the camera settings at the end of my prompt. Keep the important information at the start.

Prompt suggestions
A character of a cute dog dancing on stage, the camera is stable and fixed. 

A character of a cute dog dancing on stage, in a cinematic setting, the camera remains stable.

When I add focus on the dog, the camera may zoom in on the dog. Simple direct prompts tend to give me good results.

Cheers

Nate

Hello ​@bacizone,

Thank you for the report. The issue you are experiencing is a known limitation of AI video generation models rather than a bug specific to Firefly.

Diffusion-based video models are trained to generate content that matches descriptions rather than to suppress specific elements. Negative instructions such as "NO pan", "NO tilt", and "NO zoom" can paradoxically reinforce those concepts in the model's attention rather than eliminating them — the model essentially reads the camera movement vocabulary and associates it with the generation regardless of the "NO" qualifier preceding it.

The recommended approach is to rewrite the camera section of your prompt using entirely positive language that describes what you want rather than what you do not want. Here is a suggested rewrite of your camera instructions:

"Locked-off tripod shot. Completely static camera. Fixed frame, fixed focal length, fixed perspective throughout. Still architectural photography with subtle environmental animation only — gentle breeze moving foliage, soft light variation on surfaces, ambient atmosphere. Every architectural edge and structural element remains in identical screen position from first frame to last. Environmental motion only, zero camera motion."

The key differences are:
- "Locked-off tripod shot" replaces the list of prohibited movements
- "Fixed frame, fixed focal length, fixed perspective" replaces the zoom and push-in prohibitions
- Describing what should move (foliage, light) gives the model a dynamic element to generate without resorting to camera movement
- "Zero camera motion" is more effective than a list of "NO" instructions

It is also worth trying different video models if the issue persists — some models follow camera control instructions more reliably than others.

    droopy