Runpod

Run open-source AI models on cloud GPUs without maintaining infrastructure. Covers five serverless endpoints for image editing, upscaling, dewatermarking, avatar generation, and text-to-speech, all running on RunPod's pay-per-second GPU instances. The setup is straightforward: deploy Docker containers with a single command, and the skill handles cold starts, file transfers via R2, and polling for results. Typical jobs cost under $0.30, and the whole setup rarely breaks $10 monthly even with heavy use. The documentation includes actual GraphQL field names since RunPod disables introspection, which saves hours of API fumbling. Good fit if you need GPU workloads occasionally but don't want to babysit instances or pay for idle time.