screencastgen

TTS Base

Base classes and utilities for TTS backends.

Source: screencastgen/providers/tts/base.py

Functions

resolve_device(device="auto") -> str

Resolves "auto" to "cuda" (if available) or "cpu". Used by all GPU-capable backends.

Dataclasses

BackendArg

Defines a CLI argument for a TTS backend.

Field Type Description
name str Argument name (e.g., "--model")
type type Argument type
default any Default value
help str Help text
choices list Valid values

BackendSpec

Metadata for a TTS backend.

Field Type Description
name str Backend identifier (e.g., "qwen")
module_path str Module to import
class_name str Class to instantiate
contexts List[str] Where it can run ("local", "server")
capabilities List[str] Features ("voice_clone", "requires_ref_audio")
extra_args List[BackendArg] Additional CLI arguments
download_args List[BackendArg] Model download arguments
build_kwargs callable Extracts constructor kwargs from args
validate callable Validates args before creation
download_models callable Downloads model files

Usage Pattern

spec = get_backend_spec("qwen")
kwargs = spec.build_kwargs(args)          # Extract from argparse
spec.validate(args)                       # Check requirements
backend = create_backend("qwen", **kwargs) # Lazy import + instantiate

Dependencies

TTS Base
├── torch (deferred, for resolve_device)
└──▶ consumed by TTS Registry
     ├──▶ Qwen Backend (spec definition)
     ├──▶ Remote TTS (spec definition)
     └──▶ Lipsync Facade (resolve_device)

See Also

This site is open source. Improve this page.