Dokumentation (english)

Stable Diffusion v1.5

Latent diffusion model for generating photo-realistic images from text prompts

Stable Diffusion v1.5 generates photo-realistic images from text descriptions. Can be used with the base weights or with a fine-tuned checkpoint.

When to use:

  • Generating images from descriptive text prompts
  • Creative content and concept visualization
  • Product mockups and visual prototyping

Input: Text prompt + optional fine-tuned checkpoint Output: Generated image (PNG) and generation metadata (prompt, seed, steps, guidance)

Inference Settings

Num Inference Steps (default: 50) Number of denoising steps during image generation.

  • 20–30: Fast, lower quality — good for prototyping
  • 50: Default — good balance of quality and speed
  • 80–100: Highest quality, slower

Guidance Scale (default: 7.5) How strongly the model follows the text prompt (classifier-free guidance).

  • 1.0: Ignores the prompt, very creative/random
  • 7.5: Balanced — follows the prompt while allowing variation
  • 15+: Closely follows prompt, may produce over-saturated or distorted images

Seed (default: 42) Random seed for reproducible generation.

  • Set to a fixed value to reproduce the same image
  • Change the seed to explore different visual variations for the same prompt

On this page


Command Palette

Search for a command to run...

Schnellzugriffe
STRG + KSuche
STRG + DNachtmodus / Tagmodus
STRG + LSprache ändern

Software-Details
Kompiliert vor etwa 4 Stunden
Release: v4.0.0-production
Buildnummer: master@afa25ab
Historie: 72 Items