We built VideoRouter to reduce the cost of AI video generation. There are now many providers serving the same underlying video models, but prices can differ by 5–10× depending on the provider. VideoRouter gives you one API and routes each request across 100+ video models/providers based on price, availability, and model support. For example, the same model can be available through multiple providers at very different $/second rates. We’re building this as an inference optimization layer rather than another video-generation UI. Currently supporting providers including Atlascloud, Fal, Replicate, WaveSpeed, and others. Would be interested in feedback from people running AI video generation at scale: how are you currently choosing video generation model providers, and do you switch providers when prices change?