For the complete documentation index, see llms.txt. This page is also available as Markdown.
Starlight
Generative Video Upscale & Enhancement
Model Family Overview
Advanced diffusion based video enhancement models delivering high quality restoration of difficult archival footage and enhancement of modern GenAI video.
Generative models improve video by intelligently adding detail and texture while preserving artistic intent.
Key Features
Restore and elevate challenging video sources
Recover detail, stability, and texture from degraded archival footage and low quality sources, scaling results cleanly to 4K and beyond.
Enhance GenAI video while preserving creative intent
Predictive, diffusion-based enhancement improves clarity, consistency, and realism without altering the original visual style or structure.
Natural detail and facial realism
Refines face detail and skin tones while reducing synthetic or “plastic” artifacts common in generative video outputs.
Flexible processing: cloud or local
Starlight Mini and Starlight Sharp are specialized variants optimized for high-quality local rendering.
The above pricing is for our "Quality" models. "Fast" models are priced at 50% of the "Quality" models.
For more accurate estimates we recommend checking out our Credit Calculator.
