coqui-ai-TTS for lispeech voice like a robot

[hebo1982]

I compare coqui-ai-TTS and mozilla-TTS with lispeech models.

# Mozilla
tts --text "..." \
    --model_path tts_models--en--ljspeech--tacotron2-DCA/model_file.pth.tar \
    --config_path tts_models--en--ljspeech--tacotron2-DCA/config.json \
    --out_path speech.wav \
    --vocoder_path vocoder_models--en--ljspeech--mulitband-melgan/model_file.pth.tar \
    --vocoder_config_path vocoder_models--en--ljspeech--mulitband-melgan/config.json

# coqui-ai
tts --text "..." \
 --model_path ljspeech-dcattn-April-05-2021_01+52PM-0947822/model_file.pth.tar \
 --config_path ljspeech-dcattn-April-05-2021_01+52PM-0947822/config.json \
 --out_path speech.wav \
 --vocoder_path hifigan-April-14-2021_01+32PM-b735076/model_file.pth.tar \
 --vocoder_config_path  hifigan-April-14-2021_01+32PM-b735076/config.json  

Creates two wav files.
coqui-ai-TTS like a robot voice with a litte noise.

two-wav-files.zip

[erogol]

why --text "..." ?

and it looks like you compare different models.

[hebo1982]

The input text is too long, I replace it.

"TTS is a library for advanced Text-to-Speech generation. It's built on the latest research, was designed to achieve the best trade-off among ease-of-training, speed and quality. TTS comes with pretrained models, tools for measuring dataset quality and already used in 20+ languages for products and research projects."

And model is different.

mozilla: https://github.com/mozilla/TTS/releases/tag/v0.0.9
English | LJSpeech | TacotronDCA | tts

coqui-ai: https://github.com/coqui-ai/TTS/releases/tag/v0.0.13
English (non-binary) | sam (acccenture) | Tacotron2-DDC | tts | 😄 v0.0.13

[hebo1982]

Thanks for @erogol replay!
I replace model with "English | LJSpeech | Tacotron-DDC | tts | v0.0.12" work fine.