Crate qts 

Qwen3 TTS (GGUF + GGML) — native inference library.

Output sample rate for the published Qwen3-TTS checkpoints.

Re-exports

pub use pipeline::speaker_encoder::SpeakerEncoder;

pub use pipeline::speaker_encoder::SpeakerEncoderConfig;

pub use pipeline::tokenizer::TextTokenizer;

pub use pipeline::tokenizer::TokenizerConfig;

pub use pipeline::tts_transformer::CodePredictorConfig;

pub use pipeline::tts_transformer::CodecRollout;

pub use pipeline::tts_transformer::CodecRolloutSubTimings;

pub use pipeline::tts_transformer::IclPrefillConditioning;

pub use pipeline::tts_transformer::PrefillConditioning;

pub use pipeline::tts_transformer::PrefillForwardOutputs;

pub use pipeline::tts_transformer::PreparedPrefillInputs;

pub use pipeline::tts_transformer::SelectedCodecFrame;

pub use pipeline::tts_transformer::TtsTransformer;

pub use pipeline::tts_transformer::TtsTransformerConfig;

pub use pipeline::tts_transformer::VocoderChunk;

pub use pipeline::vocoder::Vocoder;

pub use pipeline::vocoder::VocoderConfig;

pub use pipeline::vocoder::VocoderExecutionProvider;

pub use pipeline::vocoder::VocoderGraphTemplate;

Modules

pipeline

Stage modules mirroring predict-woo/qwen3-tts.cpp (implemented incrementally).

Structs

GgufFile

ModeModelPaths

Separate mode-specific 1.7B QTS exports. All three GGUF paths may name one bundled file.

ModeSynthesizeRequest

Additive mode API: existing SynthesizeRequest literals remain compatible.

ModelPaths

Paths to the main GGUF plus vocoder artifacts.

Qwen3TtsEngine

StreamingSynthesizeResult

SynthesisStageTimings

Per-stage wall times for a single crate::Qwen3TtsEngine::synthesize_with_profile call.

SynthesizeRequest

SynthesizeResult

TensorF32

TensorI32

VoiceClonePromptV2

Enums

BackendKind

PresetSpeaker

Names only; codec ids are resolved from each checkpoint’s GGUF speaker table.

QtsMode

Qwen3-TTS checkpoint family. Modes require corresponding model weights.

Qwen3TtsError

TalkerKvMode

User-facing synthesis parameters (stable for future gdext bindings).

Constants

SAMPLE_RATE_HZ

VOICE_CLONE_PROMPT_V2_SCHEMA_VERSION

Traits

StreamingSynthesis

Functions

load_and_validate

Validate that required model artifacts exist and the main checkpoint parses as GGUF.