Poor AI generation that is still charging credits

Question

Hi, I used Firefly to generate a text to video. It was a simple promp:
People in an office gathering around a cake shaped like the letters "A" and "I." One person is taking a slice from the corner of the I.

The video generated a circular cake on a table with just hands coming in from the edges of the video and all taking a slice. No office scene. The cake wasn't shaped like AI. More than one person taking a slice. The video was nothing like the prompt, and yet there was no way to correct or refine the prompt, and generating that video still used credits. A bit mind-blowing how bad this is and how you could charge credits for something I didn't even download because it was so far from the desired end product.

Here's a screenshot of the first scene: 🤯🤦‍♂

droopydog500 · Accepted Answer

Hi @glitGRO,

Starting with an image generated with text to image and possibly refined with generative fill and using that as a keyframe can help. I did a few generations to get:

cake shaped like the letters "A" and "I." on a table with man cutting slice in an office

I then took that image to image to video and generated using your prompt.

Still some work to do, but I find it is sometimes easier to start with an image.

droopy

glitGRO · Answer

Not sure why it says :male_sign: as I didn't put any other emojis in there. 🤷🏼‍&#9794;

Sign up

To post, reply, or follow discussions, please sign in with your Adobe ID.

Sign in to Adobe Community

To post, reply, or follow discussions, please sign in with your Adobe ID.

Scanning file for viruses.

This file cannot be downloaded