AI Video Generation Service

Text-to-video and image-to-video generation service with clip concatenation and consistent multi-clip continuation. Accepts a text prompt or source image and produces video clips through multiple model backends — the API abstracts the generation model so clients submit a request and receive a video without knowing which model or node handled it.

Multi-model support with model-specific parameter mapping. Each video generation model has different input requirements, resolution constraints, and quality characteristics. The service normalises these differences behind a consistent API — same request format, same response format, regardless of backend. Job queuing with automatic dispatch to available GPU nodes based on model availability and current load.

Clip continuation maintains visual consistency across multi-clip sequences — the last frame of one clip seeds the next, producing longer coherent videos from sequential generation passes. Concatenation handles frame-rate matching and transition smoothing between clips generated on different nodes or with different parameters.

Integrates with the image generation pipeline for a complete still-to-motion workflow: generate an image, review and refine it, then animate the final version. Job tracking with full parameter provenance — every clip records its prompt, source image, model, generation parameters, and output path. Status monitoring via API for integration with automation pipelines.