Stable Diffusion 3.5 follows your prompts more closely and generates more diverse people
Stability AI has released Stable Diffusion 3.5, a significant upgrade addressing the severe quality and representation issues of its predecessor. This update prioritizes exceptional prompt adherence and diverse human representation, aiming to produce realistic, varied imagery without requiring extensive user prompting. By correcting previous failures, the new models seek to rebuild trust and establish high professional standards for open-source generation. The release includes three distinct variants tailored for different needs: a high-quality Large model for professional use, an efficient Turbo version for speed, and a consumer-friendly Medium model. These options provide developers and creators with flexible tools that balance performance with accessibility. The focus on efficiency and ease of customization ensures the technology remains viable across various hardware capabilities, expanding its potential utility. This development is crucial for open_data as it demonstrates how transparent, community-driven AI can iteratively improve upon commercial alternatives. By openly addressing past mistakes and prioritizing ethical diversity, Stability AI sets a precedent for accountability in generative AI. This reinforces the value of open-source ecosystems in fostering responsible innovation and ensuring equitable representation in artificial intelligence outputs.
Source: engadget.comPublished on 2024-10-23