Participant

In asking to generate a video from an image I explicitly told Firefly not to do something yet it goes and does it. This is not right as I've used 500 credits on a video clip I cannot use

Forum|Forum|16 days ago
March 14, 2026
2 replies
25 views

I have an image of people paddling Coracles across a river. The paddles used just have one blade, however even after telling Firefly not to use two bladed paddles in the prompt it goes and does it. I cannot use the video it produces. Your AI is just not working.

R

Ronners GaleAuthor

Participant

I've tried that passive approach several times and whatever AI I use it usually does not recognise that the initial image clearly shows a single paddle. I was told to use a negative prompt to negate the issue. It still does not work. Very frustrating and a waste of my credits.

O

Oh.N8

@Ronners Gale

Again, negative prompts do not work well.
Could you share your prompt and your reference image?
If you are using an image that has 2 paddles you will get results with 2 paddles

Use an image w one paddle, one person in a coracle.

When I go through Adobe Stock there are not a lot of coracle and a person. Rather a coracle or people and coracle.

Cheers

Nate

O

Oh.N8

Hey @Ronners Gale

Consider using positive prompts to reinforce the desired outcome.
Prompt
A person in coracle a small, lightweight, bowl-shaped boat with one paddle goes across the river

Cheers

Nate

R

Ronners GaleAuthor

Participant

The prompt image used clearly shows a coracle with one paddle. I initially used a text prompt 'using single bladed paddle'. If it cannot tell that one paddle is being used what else do you say after if fails to recognise both text prompt and a clear image. Even the word Coracle should be enough for it to work out it uses one paddle or a single bladed paddle. Wasted credits.

O

Oh.N8

@Ronners Gale

Please have a look at the images in Adobe Stock. Most of the images have multiple images of a Coracle, a person carrying a coracle, or a bunch of coracles on the beach.

Firefly is trained on Adobe Stock and other public domain images.

You may want to use a 3rd party image model. Again each model has different sources of material.

By sharing your prompt your image… we might be able to further assist you.

The more info you provide the more we can help you.
Please share your prompt and the reference image that you are using.
If the image is copyright, then we cannot really help you.

Cheers

Nate

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