Retouch Pro

26 Feb 2023

How to generate full body shots with Stable Diffusion XL

One of the common problems that AI enthusiasts experience is that AI models (such as Stable Diffusion) generate images of the upper bodies of their models – you want a full body shot, but no matter what you do with the prompt, it's always a partial body photo. We will show you how to fix this in this article.

When crafting text-to-image AI prompts to ensure generated images include full bodies, including feet, consider the following guidelines:

Detailed Description

Provide a detailed description of the scene or person you want the AI to generate. Include information about the person's appearance, clothing, surroundings, and any relevant actions or poses.

Specific Poses and Actions

Specify the pose or action of the person in the image. For full-body images, mention if the person is standing, sitting, walking, or engaged in any specific activity that requires their full body to be visible.

Environmental Context

Describe the environment in which the person is situated. This helps the AI understand the spatial relationships and ensures that the entire body, including the feet, is appropriately positioned within the scene.

Include Footwear Details

If the person in the image is wearing footwear, be sure to include details about the type of shoes or boots they have on. This will prompt the AI to include the feet in the generated image.

Clarify Perspective

Specify if the image should be viewed from a particular angle or perspective. This ensures that the AI understands how to position the person within the frame to include their entire body, including feet, from the specified viewpoint.

Provide Examples

If possible, provide reference images or sketches that illustrate the desired outcome – this can be done as LoRA or DreamBooth training of the base model, or as input images if the model permits. This can help convey your expectations more clearly to the AI model.

For example, let's try this prompt:

"A full body portrait of a young woman, (((full body))), shoulder long curly reddish hair, smiling, wearing a ((straw hat)), wearing a ((red dress))), ((dancing)), barefeet, background is italian village in the mountains, looking straight in camera, Standing on a meadow ((full of (((realistic))) flowers)) , in frame, photograph, highly detailed face, moody light, golden hour, style by Dan Winters, Russell James, Steve McCurry, centered, extremely detailed, Nikon D850, award winning photography, modelshoot style"
Try this prompt

Not ideal, we wanted a full body shot and this image doesn't even include the knees. Let's try to make the prompt simpler by removing some of the low importance keywords and also increase the height of the image slightly so that the aspect ratio is more "vertical":

"A full body portrait of a young woman, (((full body))), smiling, wearing a ((red dress))), ((dancing)), barefeet, background is italian village in the mountains, looking straight in camera, Standing on a meadow ((full of (((realistic))) flowers)) , in frame, moody light, golden hour"
Try this prompt

Nice! Now we can see model's legs and feet.