Kitten TTS API Reference

Installation

pip install soundfile
pip install https://github.com/KittenML/KittenTTS/releases/download/0.8.1/kittentts-0.8.1-py3-none-any.whl

Import

from kittentts import KittenTTS

KittenTTS Class

model = KittenTTS(model_name, backend="cpu")

Parameter	Type	Default	Description
model_name	str	---	HuggingFace model ID (e.g. "KittenML/kitten-tts-mini-0.8")
backend	str	"cpu"	Execution backend: "cpu", "cuda", "mps"

generate()

audio = model.generate(text, voice="Jasper", speed=1.0)

Parameter	Type	Default	Description
text	str	---	Text to synthesize (required)
voice	str	"Jasper"	Voice name from available_voices
speed	float	1.0	Speed multiplier (0.5 to 2.0)

Returns: numpy.ndarray -- Audio samples at 24kHz sample rate.

generate_to_file()

model.generate_to_file(text, output_path, voice="Jasper", speed=1.0)

Parameter	Type	Default	Description
text	str	---	Text to synthesize
output_path	str	---	Output WAV file path
voice	str	"Jasper"	Voice name
speed	float	1.0	Speed multiplier (0.5 to 2.0)

Returns: None -- Writes WAV file to output_path.

available_voices

voices = model.available_voices
# ['Bella', 'Jasper', 'Luna', 'Bruno', 'Rosie', 'Hugo', 'Kiki', 'Leo']

Type: list[str] -- List of available voice names.

Available Voices (8): Bella, Jasper, Luna, Bruno, Rosie, Hugo, Kiki, Leo

normalize_text()

normalized = model.normalize_text("Hello! This is text with symbols: $100 & 50%")

Parameters: text (str) -- Raw text input.
Returns: str -- Normalized text with special characters converted.

Model Variants

Model ID	Params	Size	Use Case
KittenML/kitten-tts-mini-0.8	80M	80 MB	Production quality
KittenML/kitten-tts-micro-0.8	40M	41 MB	Speed/quality balance
KittenML/kitten-tts-nano-0.8	15M	56 MB	Edge devices
KittenML/kitten-tts-nano-0.8-int8	15M	25 MB	Ultra-lightweight

Complete Example

from kittentts import KittenTTS
import soundfile as sf

# Initialize
model = KittenTTS("KittenML/kitten-tts-mini-0.8")

# Check voices
print("Voices:", model.available_voices)

# Normalize text
clean = model.normalize_text("Hello world! Test $100.")
print("Normalized:", clean)

# Generate audio
audio = model.generate(clean, voice="Bella", speed=1.0)
sf.write("output.wav", audio, 24000)

# Or generate directly to file
model.generate_to_file("Direct to file!", "direct.wav", voice="Jasper")

# Batch all voices
for v in model.available_voices:
    model.generate_to_file("Voice: " + v, v + ".wav", voice=v)
print("All done!")

Error Handling

try:
    model = KittenTTS("KittenML/kitten-tts-mini-0.8")
    model.generate_to_file("Test", "out.wav", voice="Bella")
except ValueError as e:
    print("Invalid voice:", e)
except Exception as e:
    print("Error:", e)