Getting a character head-to-toe is the most-discussed image problem in our community. The fix is usually not where people look for it.

Framing is its own prompt, not part of your description

DreamRunner adds a framing instruction to every character image — the crop and the stance — separately from your physical and outfit descriptions. It's a distinct piece of the prompt, and it's the piece that decides how much of the character you see.

That means typing "full body shot" into a character's description mostly doesn't work. You're writing into a field about what the character is, while a different part of the prompt is saying how to frame them, and the two compete.

Why we crop at mid-thigh by default

DreamRunner's default framing is a thigh-up crop — cowboy shot on tag styles, "cropped at mid-thigh" on prose ones. That is deliberate, and it is not us being precious about composition.

Diffusion models have a fixed detail budget. An image is a set number of pixels, and everything in frame shares them. The more of a subject you fit in, the fewer pixels each part of it gets. Ask for head-to-toe and the face — the part you actually look at — is rendered at a fraction of the resolution it would get in a tighter crop.

On a story card a character is only a few hundred pixels tall. Pixels spent on boots are pixels not spent on the face.

There's a second reason: consistency. Without a framing instruction the model picks the crop itself, so one character comes back full-body and the next thigh-up — and then they composite into the same scene at different scales.

But you can change it — the prompt is yours

We expose this. If you want full-body characters, you can have them.

Open Settings → Prompts, find the Image section, and edit Character Framing & Pose. There are two entries — one for tag-based image models, one for Flux and other prose models — and you edit whichever matches your style. The change is per story and applies to every character in it.

Two rules if you edit these, both learned the expensive way:

• No negations. Diffusion has no NOT; a positive prompt describes only what is there. "With the feet out of shot" reliably produces feet, and pushes the crop out to full body.
• No tag jargon in the prose version. cowboy shot means a mid-thigh crop to an SDXL model and plain English to Flux — which will cheerfully add an actual cowboy hat.

Note: going full-body is a real trade, not a free win. You'll see the whole outfit and lose detail in the face. Both are legitimate; just know which you're choosing.

Description length changes the crop too

Even with the framing prompt set, a description that is almost entirely facial detail will pull the crop tight. The model is trying to fit what you described into the frame.

If you want head-to-toe, the description needs something below the neck to be worth showing — an outfit, a stance, footwear. See How Description Weight Works.

Aspect ratio does half the work

A portrait-shaped canvas fights a full-body composition. A taller image gives the model somewhere to put a body.

Uploaded and restyled images are different

When an image is built from a photo you uploaded, DreamRunner deliberately does not impose its own framing. A restyle repaints the source, so it keeps that photo's composition — asking for a knees-up shot over a head-and-shoulders selfie demands a body that isn't in the frame, and the model fills the gap with mangled anatomy.

If you want different framing from an upload, upload a photo with the framing you want.