Multiple generations of videos without regard for my instructions.

Hi, detailed_work3e36!

I completely understand you. What I would recommend is always using a correct video/image/frame reference and remember the AI of the missing arm for that character.

I’ll leave here some tips on how to make a flawless prompt:

• Keep it simple: Firefly works best with short, easy-to-read prompts. Very long or complicated prompts (like those generated by ChatGPT) can confuse the model.

• Watch your video references: For video generation, avoid dark or busy backgrounds. The AI needs a clear contrast to distinguish the main characters from the environment!

• Copyright & Public Figures: To keep everything commercially safe, prompts requesting trademarked content, celebrities, or political figures will be blocked.

• Safety Filters (Minors & NSFW): Our safety guidelines are strictly enforced to protect minors. Firefly will not generate images of children or anything child-like. Prompts containing words like boy, girl, kid, or child will trigger the filter. Additionally, any prompts involving nudity are strictly prohibited.

Also, here are some useful links:

Generate images from text descriptions
Writing Effective Prompts
Adobe Firefly Resources

Best,

Gabriel

I am making a video that contains and android that is missing its left arm due to damage.  I have changed the cue multiple times to say that the arm is missing.  And I now have 20+ videos that the arm somehow magically reappears during the video.  These are credits that you just stole from me, and I still dont have the video I want.  At 400 credits a generation, I now feel that I am getting screwed by Adobe.