Height and Weight in Generation: Why the Model Asks
A portrait needs no build data; a full-length frame does. What happens when the fields are left empty, and when they are worth filling in.
Height and weight fields in a generation interface prompt a fair question: why does the model need dimensions when it can see a photograph? The answer is that a portrait tells it almost nothing about the body.
What a portrait shows and what it does not
A chest-up frame shows the face, the neck, the shoulder line — and there the information ends. It supports a rough guess at build, but not height, not body proportions, not how the person looks at full length.
If the scene requires showing the whole figure, the model has to invent the rest. Without data it picks the most probable option — an average build matching the face it read.
When it shows
- Full-length frames: the whole figure, head-to-body ratio, leg length.
- Waist-up frames in complex poses: seated, leaning, turned.
- Scenes where the person interacts with the environment: standing by a door, sitting at a table, walking down a street — scale is set by the surroundings.
- Frames with several people: height differences become part of the composition.
For a close portrait the parameters are not needed at all: there is no figure in frame.
What happens when you leave them empty
Nothing breaks: the model simply uses a statistically typical option. The problem appears only when the typical option diverges noticeably from reality — and then the result looks like your face on someone else's body.
This is one of those things a viewer registers instantly without knowing what is wrong: the proportions of a familiar person are off.
How to fill them in
- Real values, not aspirational ones: the point of the parameters is for the result to match reality.
- Both together: height without weight leaves build undetermined; weight without height leaves scale undetermined.
- An empty field beats a rough guess when you are unsure: the model will then honestly take a typical option rather than build a figure from wrong data.
What the parameters do not solve
They set general proportions, not individual characteristics. Posture, shoulder-to-hip ratio, body composition, fitness — all of that lies beyond two numbers.
If the figure matters in detail, supplying a full-length source photo is more reliable: it carries information no parameters can convey.
Clothing and how a figure reads
Worth remembering that clothing in the result comes from the scene and strongly affects how the figure reads. A loose coat hides proportions; fitted clothing emphasises them. So a mismatch with reality shows more in some scenes and hardly at all in others.
Privacy
Height and weight are data about your body, and it is sensible to treat them carefully. Supply them only where they genuinely affect the result, and do not fill them in "just in case": for portrait frames they add nothing, and unnecessary data is better not shared.
A practical approach: fill the fields for the specific task where height is visible, and leave them empty in every other case.
Frequently asked
Are height and weight required?
No. Without them the model uses a statistically typical build. Supply them where the figure is visible and a mismatch with reality would show.
What if I give inaccurate values?
The model builds the figure from them. If you are unsure, leave the fields empty: a typical option is safer than a figure assembled from wrong numbers.
Do I need the parameters for a portrait?
No. A chest-up frame contains no figure, and build data does not affect the result.
- #телосложение
- #параметры
- #практика