Module tts 

The [tts] section: what the speaking role runs, the Hugging Face install that supplies it, and the read the models page and the settings card draw.

Re-exports

use std::path::PathBuf; 🔒

use forge_dictate::ModelFacts; 🔒

use forge_dictate::ModelSpec; 🔒

use serde::Deserialize; 🔒

use serde::Deserialize; 🔒

use crate::dictate::DictateModel; 🔒

use crate::dictate::DictateModelState; 🔒

use crate::dictate::DictateRole; 🔒

Structs

SpeakingInstall

What one installed variant is made of, as the install recorded it.

SpeakingVariant

One ONNX variant of the model the Hub publishes.

SpeakingVariantView

One variant the page draws: whether it is installed, and whether it runs.

TtsSettings

Deliberately narrower than a model spec, for the same reason crate::dictate::DictateSettings is: a hand-edited URL, byte length or digest is how a file arrives that nothing can verify.

TtsSnapshot

The speaking role’s model rows.

TtsView

The voice read: whether [tts] is on, where the files land, the speaking model’s state, and the voices that model installed.

VoiceFile

One style file the install fetched.

Enums

SpeakingFrom

Where the active speaking model came from. A bare name on the wire: the speaking read has one from, and a tag object would nest it.

Constants

HUB_REPO

The Hub repository publishing the ONNX export and the style files the engine loads.

PINNED_VARIANT

The variant the role runs when nothing chooses another: the int8 export.

SPEAKING_MODEL

The speaking model’s own name. [tts] speaking_model accepts this, or one of VARIANTS’ names to pin a quantisation.

VARIANTS

The variants the speaking role can run, the int8 model the benchmark measured among them.

Functions

default_max_utterance_seconds 🔒