Expressive actions
How precisely do the body's expressive actions land?
Where expression tags speak with the face, these speak with the whole body — which means they still read when the face is small.
Every cut below shares one seed, one base prompt and one sampler setting, and differs only by the tag marked in amber.
What each cut shows
dancing- An arm reaches out and she looks back over her shoulder, the body curving; the cloth does not move, so it reads as a held dance pose rather than motion.
singing- The mouth opens, the eyes close, and music notes are drawn in — symbols nobody asked for.
praying- Hands together, head bowed — the most accurate of the four.
bowing- Folds at the waist. The face disappears, so it cannot be combined with an expression.
How these were made
Shared prompt
masterpiece, best quality, amazing quality, very awa, 1girl, solo, mature female, adult woman, short black hair, white blouse, blue jeans, full body, <tag slot>
Tap a tag to carry it over to the builder
Negative prompt
nsfw, nude, topless, bikini, swimsuit, lingerie, underwear, cleavage, suggestive, school uniform, pleated skirt, child, loli, teenage, bad quality, worst quality, worst detail, censor, jpeg artifacts, watermark, signature, text, extra limbs, bad hands, bad anatomy
Model
rinFlanimeIllustrious_v30 · euler_ancestral/normal · steps 28 · CFG 5 · 896×1152 · seed 777031
Takeaway
Actions that speak with the body read far more reliably than expression tags do — a contrast with how quietly this checkpoint plays a face.
How far this result reaches
Some, like singing, draw symbols into the frame. If you want a realistic image, handle them in the negative.
Measured on a single Stable Diffusion family checkpoint. The same tags can behave differently on NovelAI, Midjourney or another checkpoint. This is a reference point, not a verdict.