Struct SpeechEngine 

pub struct SpeechEngine { /* private fields */ }

The voice in memory: Kokoro 82M or LFM3 Audio, by the model’s family. Loading takes a while and speaking blocks, so keep it away from the UI thread.

Implementations

impl SpeechEngine

pub fn load( folder: &ModelFolder, model: &LocalModel, ) -> Result<Self, EngineError>

Loads model from folder. Kokoro needs model.pth, config.json and its voices; LFM3 Audio needs model.gguf, projector.gguf and tokenizer.json.

Errors

When its files are missing or cannot be read.

pub fn speak( &mut self, text: &str, settings: &SpeechSettings, go_on: &mut dyn FnMut(usize) -> bool, ) -> Result<Vec<f32>, EngineError>

Says text as the settings ask. go_on hears how much is ready and returning false stops it with what it has.

Errors

When it stops in the middle, or Kokoro is asked for a language other than Portuguese or a speaker it does not have.

pub fn pieces(&self, text: &str) -> Vec<String>

How text is best said, a piece at a time: by sentences for LFM3 Audio, and for Kokoro sentences joined while they are short, since it says short ones weakly.

Trait Implementations

impl Debug for SpeechEngine

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

Formats the value using the given formatter.