pipecrab-stt: the speech-to-text interface.
One trait — [Transcriber] — is the swappable STT capability: f32 samples
in, text out. [SttStage] adapts any Transcriber into a pipeline
Stage, turning a
DataFrame::Audio into a
DataFrame::Transcript without the
pipeline above it ever naming a concrete model.
Platform-neutral and wasm32-checkable: the concrete engines live elsewhere
(native ort, browser Transformers.js in a Web Worker), each behind this one
trait, so the interface itself carries no backend dependency and compiles for both
the host and wasm32-unknown-unknown.