Struct FixedLengthSplitter 

pub struct FixedLengthSplitter {}

No-VAD splitter: walks the waveform in bounds.max_samples()-sized strides, aligning non-final chunks to bounds.align_to_samples().

Zero model load. Suitable when the caller already segmented the input, for short utterances that fit a single chunk, or for tests. Boundary context degrades transcription quality at chunk seams — for production long-form ASR prefer SileroVadSplitter.

align_to_samples dividing max_samples is not guaranteed; the final aligned chunk is the floor of len_remaining / align_to_samples times align_to_samples (so its mel length stays an integer multiple of subsampling_factor). The very last chunk keeps its unaligned tail — the JIT pads it.

Implementations

impl FixedLengthSplitter

pub fn new() -> Self

Trait Implementations

impl Clone for FixedLengthSplitter

fn clone(&self) -> FixedLengthSplitter

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for FixedLengthSplitter

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for FixedLengthSplitter

fn default() -> FixedLengthSplitter

Returns the “default value” for a type.

impl Splitter for FixedLengthSplitter

type Error = Infallible

fn split( &mut self, waveform: &[f32], bounds: &EncoderBounds, ) -> Result<Vec<AudioChunk>, Self::Error>

fn max_chunk_samples(&self, bounds: &EncoderBounds) -> usize

Upper bound (in samples) on the longest chunk this splitter could emit. Consumed by the transcriber to size JIT buffers — a tighter advertised bound trades peak memory for tighter chunk handling. Default: full encoder capacity.