This paper presents the first large-scale empirical study of creator practices within the open-source artistic image generation ecosystem. Researchers constructed a dataset of 6 million images with embedded metadata, detailing the models and prompts used in their creation. The study analyzes the usage of 22.4K base models and 154K LoRA models, highlighting the strengths and challenges of this community-driven approach, which contrasts with closed-source tools like Midjourney. The dataset and findings are made publicly available to aid creators and researchers. AI
IMPACT Provides insights into the workflows and challenges of open-source AI model usage for artistic image generation.
RANK_REASON This is a research paper published on arXiv detailing an empirical study. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
