Small X Studio
Connecting…
Processor
Auto
GPU (WebGPU)
GPU (WebNN)
CPU (WebAssembly)
Prompt
?
Describe the image you want: subject, setting, style and lighting. Separate ideas with commas; earlier words usually carry more weight. Only the first 77 tokens (about 60 words) are used.
a cozy cabin in a snowy pine forest at dusk, warm light in the windows, highly detailed
Negative prompt
?
Things you
don’t
want in the image, such as “blurry, extra fingers, text, watermark”. The image is steered away from these at every step. Leave it blank if there’s nothing to avoid.
blurry, low quality, deformed
Steps
?
How many times the image is refined, starting from random noise. More steps add detail but take longer: time grows in proportion. 20–30 is usually enough; above 40 rarely helps.
Guidance
?
How closely the image follows the prompt. Low (1–4): looser and more varied. 7–8: balanced, the usual choice. High (12+): very literal, but colours can look harsh or oversaturated.
Width
384
512
640
768
Height
384
512
640
768
Seed (blank = random)
?
The number that sets the starting noise. The same seed with the same prompt and settings gives the same image, so you can reuse a seed you like and adjust the prompt to make variations. Each image shows its seed, with a “Reuse seed” link.
Generate
Cancel
Generated images appear here.