Stability AI provides enterprise-ready generative AI tools and solutions for content production at scale, offering image generation, editing, and 3D/4D video models.
What people report
Users are actively exploring and pushing the boundaries of Stability AI's models, particularly for video generation and LoRA training, often praising specific models like Bernini and Anima for their capabilities, while also encountering and troubleshooting common issues like model control and training efficiency.
LoRA training is a key use case
Many users are actively engaged in LoRA training for character and concept generation, experimenting with different models like KREA 2, Z-Image Turbo, and LTX 2.3. They share configurations and seek advice on optimizing training settings, dataset preparation, and captioning for better consistency.
Bernini excels in video-to-video tasks
Users are impressed with Bernini's video-to-video capabilities, especially for video inpainting, calling it the 'best and most versatile' model for this purpose. They appreciate its ability to transfer movement from video to a person in a photo, though some note challenges with face consistency.
Anima model praised for detail and hybrid prompting
Users find the Anima model delivers impressive detail, particularly in backgrounds and environments, surpassing other models like Illustrious or Pony. They also like its flexible prompting system, which combines Gelbooru-style tags with natural language.
Privacy concerns with cloud GPU and data
Users express significant privacy concerns when using cloud GPU services for ComfyUI or other AI tools, worrying about their uploaded and generated files being visible to providers or retained indefinitely. They actively seek privacy-friendly alternatives and local-first solutions.
Difficulty canceling subscriptions and generations
Users report difficulties in canceling subscriptions, with instructions often being incorrect or hidden. Additionally, many find it frustrating that they cannot instantly cancel ongoing image generations in ComfyUI, having to wait for current steps to finish, which can take minutes.