Text-to-Image
Rectified-flow diffusion in CLIP/VAE latent space.
Prompt
a red rose in a lush green garden. photorealistic
Steps
↺
1
200
Guidance scale
↺
0
10
Seed (-1 = random)
Preview every N steps (each preview adds a VAE decode, very slow). Recommended >10 for speedy generation
↺
1
200
Generate