# stream2sentence is the core of RealtimeTTS - it quickly converts streamed text into sentences for real-time synthesis
stream2sentence>=0.3.1

# azure-cognitiveservices-speech is for AzureEngine
azure-cognitiveservices-speech>=1.48.2

# cartesia is for CartesiaEngine
cartesia==2.0.9

# coqui_tts is for CoquiEngine
coqui_tts>=0.27.5

# elevenlabs is for ElevenlabsEngine
elevenlabs>=2.39.0

# gtts is for GTTSEngine
gtts>=2.5.4

# openai is for OpenAIEngine
openai>=2.29.0

# requests is for MiniMaxEngine and ModelsLabEngine
requests>=2.32.0

# pyttsx3 is for SystemEngine
pyttsx3>=2.99

# edge-tts is for EdgeEngine
edge-tts>=7.2.7

# kokoro is for KokoroEngine
kokoro>=0.9.4

# snac is for OrpheusEngine
snac>=1.2.1

# faster-qwen3-tts is for FasterQwenEngine
faster-qwen3-tts>=0.2.4

# camb-sdk is for CambEngine
camb-sdk>=1.5.8

# omnivoice is for Omnivoice
omnivoice>=0.1.1

# typecast-python is for TypecastEngine on upstream master
typecast-python>=0.2.0

# chatterbox-tts is for ChatterboxEngine
chatterbox-tts

# neutts is for NeuTTSEngine
neutts

# pocket-tts is for PocketTTSEngine
pocket-tts

# soprano-tts is for SopranoEngine
soprano-tts>=0.2.0

# sopro is for SoproTTSEngine
sopro>=1.5.0

# Shared neural engine dependencies used by ZipVoice, LuxTTS, StyleTTS, Parler, PocketTTS, and MOSS-TTS
torch
torchaudio
numpy
soundfile>=0.13.1
huggingface-hub>=0.36.0
safetensors
vocos
librosa
nltk
munch
PyYAML
phonemizer
transformers
onnxruntime
inflect
lhotse
piper_phonemize
tensorboard

# fugashi is to support japanese language for KokoroEngine
fugashi>=1.5.2
jaconv>=0.5.0
mojimoji>=0.0.13
pyopenjtalk>=0.4.1
unidic-lite>=1.0.8

# pypinyin is to support chinese language for KokoroEngine
pypinyin>=0.55.0
ordered_set>=4.1.0
jieba>=0.42.1
cn2an>=0.5.23

# pyaudio is for playing chunks over output device
pyaudio>=0.2.14

# pydub is used to convert chunks from mp3 to pcm (for openai tts)
pydub>=0.25.1

# resampy is used to resample from the tts to the target device sample rate
resampy==0.4.3
