Expand description
Shared audio workflows and engine contracts for transcription, synthesis, and voice analysis.
Re-exports§
pub use aria_primitives as primitives;
Modules§
- audio
- Audio capture: the sources, the recordings they write, and the canonical
16 kHz mono format shared by everything that reads them. A
Clipis finished audio at its own rate. - backend
- What every engine backend shares, whichever capability it serves: the boxed shape of its own failures, the files its model downloads as, and validation of the model files it loads. Each capability wraps these in its own error vocabulary.
- cancel
- Cooperative cancellation and operation deadlines.
- segmentation
- Who is speaking when: the contract every speech-segmentation engine implements.
- speaker
- Who is speaking, as a vector: the contract every speaker-embedding engine implements.
- synthesis
- The text-to-speech contract engine crates implement.
- transcription
- The speech-to-text contract engine crates implement.
- voice_
activity Non- target_family=wasm - Voice-activity inference over canonical audio frames: which stretches carry a human voice rather than music, noise, or silence.
- wake_
word