A diffusion model learns to turn random noise into a clear image step by step, guided at each step by your text prompt.
Stable Diffusion XL runs in about 10–12 GB of GPU memory; FLUX.1 gives higher quality but needs much more.
Example
A 1024×1024 SDXL image takes roughly 10–20 seconds on an L4.
বাংলায়: ডিফিউশন মডেল — লেখা থেকে ছবি বানানোর মডেল — Stable Diffusion, FLUX।