6 Million Images Reveal AI Art Ecosystem
rao2z · x · 2026-07-14
This research, "Navigating the Open-Source Model Ecosystem," analyzed 6 million Pixiv AI-tagged images to understand how creators actually use open-source image models and LoRA in the real world.
Key findings include:
- The data covers over 22,400 base models and 154,000+ LoRAs
- Despite the massive number of models, usage is highly concentrated, with 560 base models generating 80% of the samples
- By 2025, 75% of images used at least one LoRA, making LoRA the de facto standard
- Works using LoRA generally receive more views and bookmarks
- Returns diminish after stacking 3 to 5 or more LoRAs, which can even hurt the artwork's performance
- Character LoRA popularity is declining, while style/concept LoRAs are more favored because new base models can better handle direct character prompting
- Creators are cautious about upgrading to new model versions: even 20 weeks after release, only 40% of images use the latest version, partly because compatibility issues can break existing LoRAs or alter aesthetic styles
The authors have also made the dataset public on HuggingFace for others to continue analyzing prompts and metadata.
Related event: Study Analyzes 6M Pixiv AI Images to Reveal Ecosystem(2 posts)→
More from Multimodal
- Magnific teases October 28 release: 'We're making it move' — cuenca · 2026-10-27
- Midjourney --sref demo: retro Americana Batman art series — gizakdag · 2026-10-27
- AI video is a new medium, not cinema or animation — and it's perfect for depicting dreams — Kyrannio · 2026-10-27
- Dad turns his toddler's 31-year-old quote into a song via SUNO kids' choir — NathanpmYoung · 2026-10-27
- Band's local-AI music video hits 200k views: 900 generations kept 64, 3-second clips beat face drift — Hwang0500 · 2026-10-27
- Adobe Firefly video editor walkthrough: one-sentence Quick Cut turns 32s into 20s — LudovicCreator · 2026-10-26