Skip to main content
S
Star_Imaging
Known Participant
August 27, 2025
Question

AI Content Model Releases - are these required if depth or canny are used ?

  • August 27, 2025
  • 2 replies
  • 327 views

Suppose I generate AI images with a pose taken from one of my original photographs, using ControlNet, such as depth or Canny. Do I need a model release, as the created image does not resemble the original subject at all, just the pose or depth map? I would figure a release is needed when using an input image with a very low denoise value (< 0.25). With higher than 0.25 and guidance above 4.0 the final result no longer resembles the original person, so I figure I do not need a release.  I looked at the info on the Contributors site, but it is very generic regarding model releases with AI.  I have older photos where I really like the poses and styling, but Adobe does not accept the old model releases, so I was hoping to give them a new AI lease on life. Any guidance would be appreciated, as it seems to be impossible to contact Adobe anymore.

 

The example below is the resultant image

The Depth map

 The original image (this one is model released)

 

 

2 replies

Nancy OShea
Community Expert
Nancy OSheaCommunity Expert
Community Expert
August 27, 2025

I'm not familiar with ControlNet. How can you use just a pose without the original model? Aren't they one and the same?

 

 

Nancy O'Shea— Product User & Community Expert
S
Star_ImagingAuthor
Known Participant
August 27, 2025

This is what controlnet images from one of my photos look like. They no longer represent the person just the pose or depth.

Nancy OShea
Community Expert
Nancy OSheaCommunity Expert
Community Expert
August 28, 2025

Well, let me ask you this.  Would your model recognize herself in the top image if she saw it plastered on the side of a bus or building?  Do you think she'd be cool with that?

 

Nancy O'Shea— Product User & Community Expert
RALPH_L
Community Expert
RALPH_LCommunity Expert
Community Expert
August 27, 2025

 

  • If your AI-generated image is based on a real person or is intended to portray an identifiable person (e.g., you used a real person's photo as a prompt, or referenced a specific individual in the prompt), then a model release is required. The model release must be signed by the person depicted.
    This has nothing to do as to whether or not the person is identifiable. You used a real persons photo.
    It is not allowed without a release.

 

    S
    Star_ImagingAuthor
    Known Participant
    August 27, 2025

    But, then are not all images of people requiring model releases, as all AI models were trained on people to create them?

     

    I read the guideline you quoted, but the AI guidelines are vague. The image created would use a prompt that does not identify the person, and the person's image is not used as a reference. It would be an empty latent image with a generic prompt (e.g. such as a woman standing near a tree), and only the pose from a real image would be used. 

    Terms & ConditionsAccessibility statement
    Using the community Terms of use Privacy Cookie preferences Do not sell or share my personal information ad choicesAdChoices