screencastgen

TTS Registry

Central registry for TTS backend lookup, creation, and CLI/server arg registration.

Source: screencastgen/providers/tts/__init__.py

Overview

Maintains a mapping of name -> BackendSpec for all TTS backends. Creation is lazy, so backend modules are imported only when needed.

Context names in this registry are the actual parser contexts used in code:

cli
server
download for model-download-only arguments

Functions

Lookup

Function	Description
`get_backend_names(context) -> List[str]`	List backend names available in a context
`get_default_backend_name(context, preferred="qwen") -> str`	Get the default backend for a context
`get_backend_spec(name) -> BackendSpec`	Get the spec for a named backend
`iter_backend_specs(context) -> List[BackendSpec]`	Iterate specs for a context

Creation

Function	Description
`create_backend(name, **kwargs)`	Lazy-import and instantiate a backend
`create_backend_from_args(args, invocation)`	Validate args and create a backend from parsed inputs

Parser Integration

Function	Description
`register_backend_args(parser, context)`	Add backend-specific args to a parser
`register_backend_download_args(parser)`	Add model-download arguments
`get_downloadable_backend_names()`	List backends with downloadable models

Registered Backends

Name	Spec Module	Class	Contexts
`qwen`	Qwen Backend	`QwenTTS`	`cli`, `server`
`remote`	Remote TTS	`RemoteTTS`	`cli`

Dependencies

TTS Registry
├── TTS Base           (BackendSpec, BackendArg)
├── Qwen Backend       (lazy import)
├── Remote TTS         (lazy import)
└──▶ consumed by Pipeline Common
     ├──▶ CLI
     ├──▶ Models
     └──▶ Inference Server