Struct VoiceprintTracker 

pub struct VoiceprintTracker<'a, E, S = NoSegmenter> { /* private fields */ }

Available on crate feature silero and non-target_family=wasm only.

Cuts one recording into turns and judges each one. The diarizer scores whose voice a window is. The segmenter, when there is one, checks a speaker change inside a turn before the turn splits, which owner-only admission needs.

Implementations

impl<'a, E: Embedder> VoiceprintTracker<'a, E>

pub fn new(diarizer: &'a mut Diarizer<E>, at: u64) -> Self

A tracker that names enrolled voices but cannot admit only the owner. The first sample heard sits at at.

pub fn numbering(diarizer: &'a mut Diarizer<E>, at: u64) -> Self

A tracker that labels every voice, numbering those nobody enrolled. It has no segmenter, so a speaker change splits a turn unchecked.

impl<'a, E: Embedder, S: Segmenter> VoiceprintTracker<'a, E, S>

pub fn owner_only( diarizer: &'a mut Diarizer<E>, segmenter: S, at: u64, ) -> Result<Self, OwnerTrackingError>

A tracker that can admit only the owner’s words: it needs an enrolled owner, and a segmenter to check speaker changes.

Trait Implementations

impl<E: Embedder, S: Segmenter> SpeakerTracker for VoiceprintTracker<'_, E, S>

fn heard_to(&self) -> u64

Where the next sample heard sits.

fn hear(&mut self, speech: &SpeechMap, audio: &[f32])

Takes the audio after what it has heard.

fn skip_to(&mut self, speech: &SpeechMap, at: u64)

Gives up on the audio before at, which it will never hear.

fn finish(&mut self)

Judges what is still open. Call once the recording ends.

fn timeline(&self) -> &Timeline

fn forget_before(&mut self, at: u64) -> u64

Forgets the timeline before at, or before the open turn’s start when that comes first. Returns how far the speech map may be forgotten: never past what the tracker still reads from it.