Runware

Yes, for any model in the Stable Diffusion ecosystem. However, it’d be useful to know the exact model version and to double-check with tech team if it’s supported out-of-the-box as there are many versions in the market. Additionally, for ControlNets - it’s important to clarify the specific ControlNets you want to use. We support the most popular out-of-the-box (e.g. canny, depth, posing) and we can quickly add support for others - but it’s helpful to verify with us.

No, quality is completely unaffected.

Through a combination of hardware and software optimizations. Part of it is the fact that we run our custom made hardware infrastructure, with our own design of GPU board. But we also accelerate specific parts of Stable Diffusion. All these optimizations however are compatible with any SD version - there are no limitations to models or quality from our performance gains.

We should be able to achieve the same result because we can configure the setup in the same way - due to the high configuration flexibility we have. However, it’s important to know the precise parameters to achieve the same result. We can also test and iterate over the process, during integration - to ensure the results match as closely as possible.

Yes. We have pre-defined resolutions but we can add any resolution to suit your needs.

Per image - proportional to the generation time compared to the reference image. There are thousands of models and potential configurations so we don’t have a fixed priced list for any possible configuration. But the principle is that if the image takes 2x longer to generate than the reference image, it will cost 2x more. With the fastest inference times in the market, we pass these cost savings directly on to you.

We can handle pretty much any scale and currently support customers generating tens of millions of images per day with dedicated capacity. Get in touch to learn more.

We do not restrict NSFW generation - therefore this depends on the model you select and how it was trained. We allow the running of any SD model, however we do provide a NSFW detection filter - which adds a flag to the image if NSFW is detected. This is the standard filter created by Stability. It can have both false positives and false negatives - but it is currently the industry standard. Generally, we do not limit content generation in any way. But we of course expect that all content generated will be in legal guidelines.