Add positives like “level horizon, straight camera” and if it still persists, use negatives like “tilted horizon, dutch angle”. This is one of the few camera-style issues where a negative can be a clean fix if it keeps repeating. Also avoid cinematic keywords that imply stylized angles. If you want cinematic but level, specify “cinematic framing, level horizon” together.

Use: “bathroom mirror selfie, vanity lights, face centered, eyes in focus, sharp focus.” Vanity lights imply bright frontal light, which helps clarity.

“commercial product photo, studio lighting, seamless background, sharp focus, controlled highlights, realistic reflections, clean edges.” Append the material: “metal/glass/wood.” This forces a geometry-and-lighting prior that reduces mush and improves shape fidelity.

It can, depending on how you set it up. Many wizards output a hybrid: tag-like tokens for style + SDXL photo language for realism control.

If password reset is available, it is the only legitimate reason we’d send a reset link. We will never send you a password in plaintext. If you didn’t request it, ignore it and consider changing your password.

Describe one simple gesture precisely: “right hand holding mug handle” or “peace sign with right hand.” Keep the camera close enough that the hand has pixels, and keep lighting soft so finger edges aren’t harsh. Add a short negative: “extra fingers, deformed hands.” Then inpaint if needed.

Pixel budget. Hands are tiny in full-body shots, so the model guesses details. Fix: 1) Increase resolution and/or choose a crop where hands are larger (mid-shot). 2) Choose simpler hand poses (hands not doing anything). 3) Avoid props in hands (phones, cups) until hands are stable. 4) Use “hands visible” only if you can actually give them pixels; otherwise keep them out of frame. 5) Use a short negative block for hands. 6) If you need perfect hands, compose specifically for them—don’t expect perfect fingers in a tiny full-body frame.

Because the subject details are spread across more area. The face gets fewer pixels. Increase resolution or switch to waist-up. If you need full-body, keep the background simple and reduce pose complexity. Full-body is a harder problem; don’t compare it to headshots and assume the model “got worse”.

It encourages background defocus and reduces clutter. It pairs well with “portrait” and “bokeh,” and often increases perceived quality.

Because they can be used for impersonation, harassment, fraud, or non-consensual content. Platforms often restrict them to reduce harm and legal risk.

Midday sun creates top-down harsh shadows. Prompt it honestly: “midday sun, hard shadows,” then add a fix: “reflector fill, softened shadows under eyes.” Real photographers use reflectors or open shade—prompt “open shade” if you want flattering outdoor light: “subject in open shade, bright background.” Without fill/open shade cues, SDXL will produce harsh eye sockets that look unflattering and sometimes uncanny.

Use: “bathroom mirror selfie, vanity lights, minimal glare, face centered, eyes in focus, sharp focus.” Add “clean mirror” and “no smudges” if reflections get messy.

Not really. Tags describe what you want; negatives help block what you don’t want (artifacts, bad anatomy, text). They work together.

Use a fixed seed, keep the same server/type, avoid changing sampler/scheduler, and don’t add/remove heavy prompt elements between runs.

If you push heavy distortion + low-res, details suffer. Keep the subject larger in frame and don’t overdo compression tags.

Rapid location hopping, many logins in short windows, multiple accounts from the same device/IP, and patterns that resemble bots or shared accounts.