How SD process the strength parameter text such as "(amphibic:1.2)"?

Stability-AI / stablediffusion

High-Resolution Image Synthesis with Latent Diffusion Models

MIT License

37.84k stars 4.88k forks source link

How SD process the strength parameter text such as "(amphibic:1.2)"? #346

Open chg0901 opened 7 months ago

chg0901 commented 7 months ago

I want to discuss an issue. In the text prompts of Stable Diffusion, there are often many strength parameters. For such prompt texts like ‘fractal art:1.2’, ‘fantasy style (amphibic:1.2)’, do they have a dedicated text encoder? How is it processed?

If there are similar prompt requirements (adding strength parameters), how should the model be trained or adjusted, or how should the training data be prepared?

chg0901 commented 7 months ago

@mrorii @palp @rvencu @apolinario could you please give some hints? Thank you sir~