Speech Synthesis Factory
Author: Kongdw
kongdw8@gmail.com
dy2009@126.com
Eagle-TTS
Eagle-TTS: A Hybrid Architecture for Real-Time, Human-Indistinguishable Speech Synthesis
Integrating Llama, Parallel Attention and Conditional Flow Matching
Kongdw
1, data_io format
id=cur_id|text=*|wav=cur_id.wav|phone=*|phone_npy=*|spk_embedding=*|whisper_feat_medium=*
2, tradition tts, parallel
vits,fast-speech,matcha-tts,f5-tts
3, tradition tts, rnn
tacotron,
4, gpt,Llama based TTS
cosyvoice2
5, vocoder
hiftnet
train:
python train.py
test:
python infer_01.py