gmum/beta-CFG
This paper presents β-CFG, a dynamic guidance method for text-to-image diffusion models. Unlike standard CFG, which uses a fixed guidance scale, β-CFG adapts guidance strength over time using a β-distribution. This improves image quality, keeps sampling closer to the data manifold, and achieves better FID while maintaining prompt alignment.
When creating images from text descriptions using AI, there's a common struggle: either the image looks high-quality but doesn't quite match your prompt, or it perfectly matches the prompt but looks less artistic. This tool helps fine-tune the balance, dynamically adjusting how strictly the AI follows your text prompt. It takes your text prompt and generates a higher-quality image that maintains strong relevance to your original description. This is for AI artists, designers, or anyone generating visual content from text.
No commits in the last 6 months.
Use this if you are generating images from text prompts and want to improve the overall quality of the generated image while ensuring it still accurately represents your text description.
Not ideal if you need a tool for basic image editing, photo manipulation, or generating images without any text input.
Stars
10
Forks
—
Language
Python
License
—
Category
Last pushed
Mar 02, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/gmum/beta-CFG"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
UCSC-VLAA/story-iter
[ICLR 2026] A Training-free Iterative Framework for Long Story Visualization
PaddlePaddle/PaddleMIX
Paddle Multimodal Integration and eXploration, supporting mainstream multi-modal tasks,...
keivalya/mini-vla
a minimal, beginner-friendly VLA to show how robot policies can fuse images, text, and states to...
adobe-research/custom-diffusion
Custom Diffusion: Multi-Concept Customization of Text-to-Image Diffusion (CVPR 2023)
byliutao/1Prompt1Story
🔥ICLR 2025 (Spotlight) One-Prompt-One-Story: Free-Lunch Consistent Text-to-Image Generation...