Anima preview3 was released

Circlestone Labs has launched the third preview version of its AI image generation model, Anima preview3, marking another step in the model's ongoing development. The release, shared on platforms like Hugging Face and Civitai approximately two hours ago, is explicitly noted as a work-in-progress, with training still actively underway. This iterative approach allows the community to test and provide feedback on incremental improvements.

The creator highlights two major technical advancements in this version. First, the model has undergone significantly longer training at a 1024-pixel resolution, a process labeled 'Highres training' that is crucial for generating detailed and coherent larger images. Second, and perhaps more impactful for artistic diversity, the team has expanded the model's dataset. This expansion specifically targets 'less common artists'—defined as those with an online footprint of roughly 50 to 100 posts—which should help the model learn and replicate a wider, more nuanced range of artistic styles beyond the most popular ones.

This release underscores the rapid, community-driven pace of open-source AI model development. By publicly releasing previews, Circlestone Labs engages a global pool of testers who can stress-test the model on Civitai and Hugging Face, providing invaluable real-world data to guide future training cycles. The focus on higher resolution and niche artist inclusion directly addresses common user requests for more control and variety in AI-generated artwork.