search

coqui-ai / TTS

🐸💬 - a deep learning toolkit for Text-to-Speech, battle-tested in research and production
http://coqui.ai
Mozilla Public License 2.0
34.69k stars 4.21k forks source link

tts-server: add dropdown to select model #737

Closed mariusa closed 3 years ago

mariusa commented 3 years ago

Describe the bug In the docs screenshot there's a dropdown to select desired model: https://github.com/coqui-ai/TTS/raw/main/images/demo_server.gif

but in local tts-server there's no dropdown: 2021-08-17_00-00

 tts --list_models
 Name format: type/language/dataset/model
 1: tts_models/en/ek1/tacotron2
 2: tts_models/en/ljspeech/tacotron2-DDC [already downloaded]
 3: tts_models/en/ljspeech/tacotron2-DDC_ph
 4: tts_models/en/ljspeech/glow-tts
 5: tts_models/en/ljspeech/tacotron2-DCA
 6: tts_models/en/ljspeech/speedy-speech-wn
 7: tts_models/en/ljspeech/vits
 8: tts_models/en/vctk/sc-glow-tts
 9: tts_models/en/vctk/vits
 10: tts_models/en/sam/tacotron-DDC
 11: tts_models/es/mai/tacotron2-DDC
 12: tts_models/fr/mai/tacotron2-DDC
 13: tts_models/zh-CN/baker/tacotron2-DDC-GST
 14: tts_models/nl/mai/tacotron2-DDC
 15: tts_models/de/thorsten/tacotron2-DCA
 16: tts_models/ja/kokoro/tacotron2-DDC
 17: vocoder_models/universal/libri-tts/wavegrad
 18: vocoder_models/universal/libri-tts/fullband-melgan
 19: vocoder_models/en/ek1/wavegrad
 20: vocoder_models/en/ljspeech/multiband-melgan
 21: vocoder_models/en/ljspeech/hifigan_v2 [already downloaded]
 22: vocoder_models/en/ljspeech/univnet
 23: vocoder_models/en/vctk/hifigan_v2
 24: vocoder_models/en/sam/hifigan_v2
 25: vocoder_models/nl/mai/parallel-wavegan
 26: vocoder_models/de/thorsten/wavegrad
 27: vocoder_models/de/thorsten/fullband-melgan
 28: vocoder_models/ja/kokoro/hifigan_v1

If the bug cause is having only 1 model downloaded, I'd expect for all models to be listed and download on demand.

To Reproduce

  1. $ pip install TTS
  2. $ tts-server
  3. open provided URL in browser. Generating speech works, but there's no dropdown to select model.

Expected behavior Dropdown to choose from all possible models, even if not downloaded yet.

Environment (please complete the following information):

Additional context Since this is a local demo to try out Coqui, please have larger default limits. Now generated speech stops at 6 seconds Decoder stopped with `max_decoder_steps` 500

Please allow 10 min by default.

erogol commented 3 years ago

It is not a bug. It is the way it is intentionally not to bust people's computers with all these models. Maybe you can rephrase it to be a feature request in which I guess you want to download models on demand from the server UI.

stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discussion channels.