CLIP-txt2img-diffusers-scripts

Example scripts for using my fine-tuned CLIP models with HuggingFace Transformers / Diffusors Pipeline. 🤗

➡️ CLIP-L_and_LongCLIP-TE-Flux.py

• Use OpenAI/CLIP-L vs. My CLIP-L and beichenzbc/Long-CLIP vs. My Long-CLIP

• ⚠️ EXAMPLE SCRIPT: 'How to use CLIP' with Flux.1. Not optimized.

• ⚠️ Insert your own optimal Flux Pipeline / config for best performance!

• ✨ Pro TIP 👀 (especially [short] CLIP-L)! Hide the text-for-in-image from CLIP. If prompt 'some text here':

• Long my prompt yadda [...] long yadda | Truncate due to >77 tokens | token 78++ is: holding a sign that says 'some lengthy text'

• Or my normal prompt | PAD, PAD, PAD, PAD | token 78++ is: holding a sign that says 'some lengthy text'

• Result: Better image due to better CLIP-L guidance. Better TEXT in image due to hiding that part of the prompt from CLIP!