Copy link to clipboard
Copied
I attempted to generate video from two different photos of a man and his mother and a family walking in the woods. Both were immediately rejected. Then I used Firefly to generate an image of a man holding his daughter and tried to use that, but it was also rejected. I've attached these images here. I can find no reason that the system would reject these on a content-level. So any thoughts you have would be appreciated.
Hello @Trace35191221nizm,
Thank you for your message. I am sorry you are having this problem.
The beta version of image to video will not generate videos from images containing children. That is why it would not generate the videos.
My best,
droopy
Hello @Suchanda ,
At this time, the model will not generate videos based on keyframes that contain children. That is a limitation of the model.
droopy
Hi @Felix3LGato !
Thanks for pointing that out. Firefly currently avoids generating videos from images that include children. This is a careful step to make sure the model handles such content responsibly. We know it can be limiting, but it’s part of ensuring safety as the tool improves.
Thanks
Copy link to clipboard
Copied
Hello @stefanb89086155,
Firefly video currently has a restriction that it will not generate a video where a keyframe has what it believes is a depiction of a child. That restriction is regardless of the origin of the image used in the keyframe.
My best,
droopy
Copy link to clipboard
Copied
Copy link to clipboard
Copied
Hello @Med Lyn,
The issue most likely is that currently, image to video will not generate videos from images that depict what it believes are children.
My best,
droopy
Copy link to clipboard
Copied
I submitted this picture as a starting frame and gave it the prompt of "Have everyone slowly shifting around and looking at each other playfully and lovingly as they line up for a photo shoot" and your system tells me I'm violating a user guideline.
In this day and age of AI.. where it's likely an AI that's deciding I've violated a guideline, wouldn't it help if your system told me WHAT I'd done wrong instead of "read these several reams of webpages and guess"?
Where did I go wrong?
Copy link to clipboard
Copied
Hello @Allison37192506siip,
At this time, image to video will not generate videos from keyframes that contain children. That is a current limitation of Firefly video.
droopy
Copy link to clipboard
Copied
so should the guidelines be updated from " [Can not]....Depict nude minors or minors in a s*xual manner"
To "Cannot depict minors in any manner"?
So that they reflect the actual rules?
[Update: lol, I can't actually quote the guidelines here because of word censoring..]
Copy link to clipboard
Copied
You're automation reviewing if images meet the guideline are inaccurately blocking an image that meets the shared guidelines.
Copy link to clipboard
Copied
Hello @Scott22039795jleq,
In image to video, the model will not use as a keyframe an image that depicts children. That is a current limitation of Firefly video.
My best,
droopy
Copy link to clipboard
Copied
Hola, estoy teniendo un problema con la app de imagen a video de firefly, al insertar una imagen generada por IA desde la opción de texto a imagen, me genera el siguiente error:
Copy link to clipboard
Copied
Hello @Carlos34243557974m,
In image to video, the model will not use as a keyframe an image that depicts children. That is a current limitation of Firefly video.
My best,
droopy
Copy link to clipboard
Copied
Thank you so much for your quick response! is there any other choice via adobe video? because if i do it via text to video it works... i need to generate some stuff for some clients... nothing bad at all any suggestion?
Copy link to clipboard
Copied
Yeah, it is only a restriction in image to video. I tried to use the Veo 2 model and it seems to have the same restriction. It is not clear whether this is a temporary or pemanent restriction, but in any case for now there does not seem to be any way around this issue.
droopy
Copy link to clipboard
Copied
I just subscribed to Firefly to be able to turn images into quick videos. It continually tells me my images go against their guidelines. I read the guidelines and see nothing I am doing wrong. Pictures like me at the age of 5 outisde in the yard with my Uncle, or me with my 2 sisters, or one of me and my husband. Nothing xrated or even PG rated. I ask for simple movements, and it just keeps rejecting them all. Out of the 5 or 6 tries, its only accepted 1 of my images. In frustration, I tried to cancel my subscription but its already been paid for so I guess I may as well keep trying for the next month and then cancel it.
Copy link to clipboard
Copied
Hello @sshalimar16,
Image to video currently has a restriction that it will not generate videos where the keyframe contains what it believes are children. While that is not explicitly called out in the Generative AI guidelines, it is a restriction in Firefly Video. Adobe has not said if that is a permanent restriction.
My best,
droopy
Copy link to clipboard
Copied
I created an image on firefly. It is a simple image of an ancient family in the desert. It shows a man wearing an ancient robe, a woman with a headcovering, and two boys.
I imported this image into Firefly generative video, and it refused to use the image, citing it as an offensive image against the guidelines. Like, What?
Firefly refused to use an image created by Firefly!
I will attach the "offensive" image created by Firefly.
Copy link to clipboard
Copied
Hello @christopher7,
I merged your message into this thread which covers the issue you are having. The current Firefly Video model will not generate videos based on keyframes that contain children. That is a current limitation of the model.
My best,
droopy
Copy link to clipboard
Copied
So I can't use an AI generated image of children, made by Firefly itself to animate it on Firefly?
I can't image this being a problem. Is this limitation a bug, or a deliberate limitation by design?
Copy link to clipboard
Copied
The source or style of the image does not matter. I believe this is out of an abundance of caution to reduce the potential of generation of videos that represent children in inappropriate ways. I do not speak for Adobe. I get the sense that this will not necessarily be a permanent restriction and more internal testing is needed, but I do not have any specific information on whether or when this might change.
droopy
Copy link to clipboard
Copied
The problem with this argument, only imho, is that 1) If adobe's AI is intelligent enough to identify a child within an image, meaning it has to analyze their body, fill in where they're blocked, and develop a 3-dimensional profile for it - then it's intelligent enough to recognize misuse of that image. And, 2) how would restricting images not allow them to do the testing they need? I'm hoping they're not in their labs creating unlawful images of kids?? By allowing the community to participate, they could have the AI trained in days.
Copy link to clipboard
Copied
Copy link to clipboard
Copied
I merged your message into this thread which covers the issue you are having. The current Firefly Video model will not generate videos based on keyframes that contain children. That is a current limitation of the model.
My best,
droopy
Copy link to clipboard
Copied
I see in forums that the beta version of firefly didn't allow children in reference frame photo is that still the case for the full version as I assume this is no longer in beta?
Copy link to clipboard
Copied
Copy link to clipboard
Copied
Copy link to clipboard
Copied
Hello @Suchanda ,
At this time, the model will not generate videos based on keyframes that contain children. That is a limitation of the model.
droopy
Find more inspiration, events, and resources on the new Adobe Community
Explore Now