Struct CeltDecoder 

pub struct CeltDecoder { /* private fields */ }

Implementations

impl CeltDecoder

pub fn new(mode: &'static CeltMode, channels: usize) -> Self

pub fn seed_from(&mut self, src: &CeltDecoder)

Seed this decoder’s inter-frame state from another decoder (typically the auxiliary mono decoder), replicating its channel 0 into every channel of self. Used at a mono->stereo switch so the primary stereo CeltDecoder’s overlap/energy/prefilter state is continuous with the preceding mono packets (which libopus keeps in one continuous decoder) — without this the first stereo frame’s MDCT overlap-add starts from silence.

pub fn set_stream_channels(&mut self, sc: usize)

Channels coded in the next packet’s bitstream (1 for a mono packet decoded by a stereo decoder — keeps 2-channel state continuous across switches).

pub fn reset(&mut self)

libopus OPUS_RESET_STATE for the decoder: clear everything from rng onward, then oldLogE/oldLogE2 = -28 (oldBandE stays 0).

pub fn decode( &mut self, compressed: &[u8], frame_size: usize, pcm: &mut [f32], ) -> usize

pub fn decode_with_start_band( &mut self, compressed: &[u8], frame_size: usize, pcm: &mut [f32], start_band: usize, ) -> usize

pub fn decode_from_range_coder( &mut self, rc: &mut RangeCoder, total_bits: i32, frame_size: usize, pcm: &mut [f32], start_band: usize, ) -> usize

pub fn decode_from_range_coder_with_band_range( &mut self, rc: &mut RangeCoder, total_bits: i32, frame_size: usize, pcm: &mut [f32], start_band: usize, end_band: usize, ) -> usize

pub fn conceal_lost(&mut self, frame_size: usize, pcm: &mut [f32])

Packet-loss concealment for a lost CELT frame — a port of libopus celt_decode_lost (celt_decoder.c). For the first few losses of a burst it uses the pitch-based branch (LPC-whitened excitation extrapolated at the last pitch period, resynthesized through the LPC filter — good for tonal/music content); once the burst runs long (loss_count >= 5) it falls back to the noise-based branch (spectrally-shaped, energy-decayed random excitation). Both fill the decode buffer, then this deemphasises to pcm (interleaved, /32768). Real attenuating audio instead of silence.