Struct SynthesizedAudio 

pub struct SynthesizedAudio {
    pub frames: u64,
    pub prepared_token_count: usize,
    pub pcm: Vec<f32>,
    pub ttfa: Option<Duration>,
}

A completed synthesis: the codes the talker produced and the audio they decode to.

Fields

frames: u64

Codec frames generated before the stop.

prepared_token_count: usize

Token ids that entered the model path, including the assistant wrapper.

pcm: Vec<f32>

Mono 24 kHz samples in [-1, 1].

ttfa: Option<Duration>

Time from synthesis start (prompt work + prefill + first frames) to the first decoded packet of PCM existing. None when the run produced no audio. Time-to-first-audio and real-time factor are different products (doctrine: report them separately); this is the TTFA half, excluding model load, which the load stage event already bounds.