Stability AI - the practical guide.

Stability AI is the company behind the open-weight Stable Diffusion image models, plus newer Stable Video, Stable Audio and Stable LM model families. It offers paid API access via the Stability platform (and partners like Amazon Bedrock and AWS Marketplace) and continues to release open weights that the wider community fine-tunes and self-hosts.

Developers and enterprises building generative image, video or audio features into their own products, plus teams that want to self-host open-weight models for cost or data-residency reasons. Less of a fit for non-technical end users who just want a polished consumer chat app.

API pricing is per generation, with different rates for SD 3.5 Large, Stable Image Ultra, Stable Video, etc. Self-hosted use of open weights is free, with infrastructure and GPU costs your responsibility. Enterprise licensing is available with indemnification and dedicated support.

Right when you want to embed image/video generation into your own product, control costs at scale, or run models on your own infrastructure. Wrong fit when you want a turnkey UI for end-users (use Midjourney/Ideogram/Firefly) or when you need the absolute top of consumer image quality without engineering effort.

Open vs commercial licensing terms have shifted over time - re-check the licence for any model you ship in a product. Self-hosting requires real GPU and MLOps investment. Latest Stable Diffusion versions have closed the gap with rivals but don't always lead on specific styles.