Guide to character posing and how to get prompts right - Solo Character Linda

First of the series that looks at mastering prompts and we start with using solo characters.

Getting precise, natural-looking character poses in AI image generation often requires a trial-and-error process of refining prompts and using reference images. When creating keyframes for short videos or specific storyboard scenes, getting body positioning and camera angles right is the difference between an average generation and a flawless final render.
Using a fictional test character created in Mage Space, this step-by-step guide demonstrates how to evolve a basic, unspecific prompt into a perfectly posed image. By iteratively adjusting camera directions, adding physical context, and leveraging past generations as reference images, you will learn how to guide the AI to respect complex poses without confusing the model.
In this guide:
  • Creating a Character through Apps - Meet Linda: Generating a consistent base character using Mage Apps
  • First Pose: Analyzing how AI interprets vague, unspecific pose instructions
  • Direction and adding context: Specifying body orientation and motion cues to prevent awkward postures
  • Adding more details: Using previous generations as reference images to lock in scene context
  • Achieving perfection: Refining limb positions and avoiding contradictory prompt statements
Because I didn't want to use a real life example, I created a fictional person through the Apps section on Mage, and created a character that I will call Linda. Mage is great at creating hypothetical fictional people with realistic skin tones and photo-realistic look and feels.

Creating a Character through Apps - Meet Linda

After answering questions relating to skin tone, hair colour/length, clothing, age, etc, I was presented with my character Linda, who I subsequently added to my general list of Characters that I will draw from.

Note that Mage created all the additional aesthetics and I had no choice in things like the clothing specifics, hair type (except that its long and wavy), her face, her pose and the background. All the background and clothing aesthetics can be modified using the Image modifier or by applying references objects like clothing
My first prompt scenario will be that Linda will be on all fours and looking for her contact lenses.

in my mind, I imagine she is perfectly on all fours on the ground, hands and knees flat on the floor and looking downwards. My first prompt is like this.

First Pose

Note that this is an extremely lazy way of prompting and I'm really giving the AI not much information to go on.

Though I used the character @mention to retrieve the Linda character, I also didn't provide any kind of additional information or references that would change the background, settings, clothing, etc. Because of this, the AI defaults to the Character image saved and uses the defaults.

The result is not quite what I expected, she absolutely looks like she is praying or something. Very weird, but perhaps the AI interpreted "find her contact lens" with this kind of sitting position. I also probably expected a side view to the left, and she is wrongly positioned even though I didn't mention any specifics. The AI fills in any information not provided, and this is exactly why the unlimited model of image generation is best for these types of AI creations.
Refining this prompt, I add more context in hopes that the AI will understand what I'm trying to create.

I add that she is facing left, and also that she is "crawling around" as an action so that the AI will understand that she is moving and not doing a "sit kneel" position.

Direction and adding context

She no longer looks like she is praying, but at least she is facing left. Strangely enough her eyes are no longer looking at the ground in search of the contact lens, and it seems like a glitch. I will keep refining the image to get it to what I want.
So that i don't lose the details of the previous prompt, I now add the previous image as a reference image to the prompt so that it has a good starting point.

Whenever I get frustrated with a picture, I find it useful to stop, save the last good image and then write the next prompt from that point onwards. In this way, I do not need to reapply past prompt and direction cues to the image as these requirements are already embedded in the underlying image.

My next prompt is more of a refinement and improvement rather than complete instructions. I also mention additional changes to help drive home what I'm after.

Adding more details

Because I've added the last statement, I've told the AI what her lower torso position should be and didn't need to restate the rest of her body movements.

Oddly enough, her shoes were removed too. Didn't ask for it, but it just figured that this position would require that. Typically I would need to add something like "keep her shoes on" at the end of the prompt.

For me, the image isn't quite perfection yet. I've got one more change to add.
Because I want the position to be a pure "all fours" position on the floor, I make sure to add "her arms not bent" at the end to have the AI make this change.

If you are doing poses like this, make sure not to provide contractionary statements in your prompt, otherwise it will get confused. If you make mention about limb or body movements once, don't restate additional instructions, it just can't handle more than one order.

I've also added the last good image as the main reference image as well so that the AI doesn't have to struggle.

Achieving perfection

The result is exactly what I wanted.

She is facing the correct position, she is not in the sitting/kneeling position, she is truly on all fours and limbs are correctly posed. Oddly enough, her shoes came back on.

Not sure if she found her contact lens, but I found perfection.
I make a lot of mistakes since I'm making this as a demonstration. But typically if you have the completely prompt correctly in your mind and written down, you can probably achieve your own "perfection" in the one image generation without all the errors and re-running of images.

However, what words and cues does Mage's AI truly respect? To understand this, you will need to go through a lot of trial and error, and maybe get the help of the Prompt helper

Conclusion

Q: What is a solo character prompt in AI image generation?

A solo character prompt focuses on generating a single consistent fictional person (like Linda) across different scenes, maintaining consistent features like face, hair, and style while varying actions or backgrounds.

Q: Why does the AI generate awkward or unexpected poses when given simple instructions?

When given vague or lazy instructions (e.g., just describing an action without orientation or context), the AI fills in the missing information based on its default training weights, often resulting in misinterpretations like strange sitting or kneeling positions.

Q: How can I stop the AI from misinterpreting body positions and actions?

You can fix this by being specific with your direction and context. Instead of just stating an action, add clear body orientation cues (such as facing left or right) and precise motion descriptions to guide the model.

Q: What is the benefit of using an iterative prompting workflow?

An iterative approach allows you to build upon past generations. By refining camera directions, adding physical context, and using previous outputs as references, you gradually train the model to respect complex poses without confusion.

Q: Can I keep my character's appearance consistent while changing their pose?

Yes, by referencing your base character (such as using a character mention tag or a saved anchor image) and layering precise pose modifiers, you can alter their posture while preserving their core look.

AI character pose tutorial

This video is relevant because it demonstrates advanced techniques for controlling AI character poses and keeping consistency across generations.

Frequently Asked Questions (FAQs)