Inference Methods

infer()

Generate speech from a single text input.

audio = tts.infer(
    text: str,
    ref_audio: str = None,
    ref_codes: Tensor = None,
    ref_text: str = None,
    voice: dict = None,
    max_chars: int = 256,
    silence_p: float = 0.15,
    crossfade_p: float = 0.0,
    temperature: float = 1.0,
    top_k: int = 50,
    skip_normalize: bool = False,
)

Parameters

Parameter	Type	Description
text	str	Text to synthesize
ref_audio	str	Path to reference audio for voice cloning
ref_codes	Tensor	Pre-encoded reference codes
ref_text	str	Transcript of reference audio
voice	dict	Preset voice dict from get_preset_voice()
max_chars	int	Max characters per chunk (default 256)
silence_p	float	Silence duration between chunks in seconds
crossfade_p	float	Crossfade duration between chunks
temperature	float	Sampling temperature (higher = more variation)
top_k	int	Top-k sampling (limits token choices)
skip_normalize	bool	Skip text normalization

Returns

numpy.ndarray — Audio waveform at 24 kHz.

Voice Priority

1. voice dict (from preset)
2. ref_audio + ref_text (voice cloning)
3. ref_codes + ref_text (pre-encoded)
4. Default preset voice (if available)

---

infer_batch()

Generate speech for multiple texts.

audios = tts.infer_batch(
    texts: List[str],
    # Same voice/sampling params as infer()
)

Returns

List[numpy.ndarray] — List of audio waveforms.

PyTorch mode uses true batch generation. GGUF processes sequentially.

---

infer_stream()

Generate speech with streaming output (GGUF only).

for chunk in tts.infer_stream(
    text: str,
    # Same voice/sampling params as infer()
):
    play_audio(chunk)

Yields

numpy.ndarray — Audio chunks as they're generated.

---

save()

Save audio to a WAV file.

tts.save(audio: numpy.ndarray, output_path: str)

---

encode_reference()

Encode reference audio to reusable codes.

codes = tts.encode_reference(ref_audio_path: str)

Returns

torch.Tensor — Encoded speech codes.

---

close()

Release model resources and free memory.

tts.close()

Or use the context manager:

with Vieneu() as tts:
    audio = tts.infer(text="...")