Skip to main content

Model Adapters and Options

Rilo uses adapter modules to normalize model-specific inputs:

  • text-to-image adapters
  • image-to-video adapters

This allows swapping models without changing orchestration logic.

How options are applied​

  • models selects which provider model is used per category.
  • modelOptions is resolved against the selected model metadata.
  • Unknown keys, invalid types, and out-of-range values are rejected.

Categories​

  • textToText: script generation model options
  • textToSpeech: voice and speech controls
  • textToImage: keyframe generation options
  • imageTextToVideo: segment generation options
  1. Start with default models and no overrides.
  2. Add one option at a time in modelOptions.
  3. Run a short project and inspect output quality and runtime.
  4. Save stable presets in project config for repeatability.

See: