Thoughts about new CLI (WIP ; will PR soon)

[p0p4k]

I found TTS command line interface to be slightly un-intuitive ; be it typing complete model names OR re-loading models for every "--text" run even on the same model. On top of it, typing --"parameter" "value" seems so outdated. So, just to spice some things up, introducing a better and modern CLI.

• It is a WIP.
• Based on python-inquirer package.
• My plan is to add all the features from existing CLI (with the help of synthesize.py).
• Supports holding the loaded model in memory to avoid re-loading it again unless "tts restarted".
• Supports changing speaker_idx and other multi-speaker params for the same loaded model without re-loading it.
• Gonna update the description as I add more features.
• Please comment if you have any new features in your mind.
  

[lexkoro]

Will running the CLI in a bash script still be possible?

[p0p4k]

What does running in a bash script actually mean? I am confused.
Right now, you can type "tts-cli" in cmd/shell and run the interface. If you write something like "python inquire.py" in the script.sh , it will run the interface as well. But, then do you want to automate the inputs, for testing and such? Is that the intent? Thanks.

[lexkoro]

Some people have written bash scripts where they process large portions of text in a bash script and synthesize each sentence separately for example using the tts command in a loop.

Example: #1622

Also yeah sorry, the question was formulated kinda unclear.

[p0p4k]

@lexkoro Hope you had a good weekend.
After giving it some thought, I think in that case we keep both 'tts' (existing cli) and 'tts-cli' (this module),
'tts' can be used for auto-scripts, tests, etc., while 'tts-cli' might be useful for complete beginners who would want an easier graphical experience (?).
Either way, it is still in early stages and if I can find a way to integrate 'tts-cli' for bash scripts, I will update here. Thanks for the suggestion.

[erogol]

I think instead of selecting something from a long list of models, it is easier to write its ID.

Something like

$ Choose a TTS model (type model ID): 4

[p0p4k]

Here, we are typing nothing, just using arrow keys to choose the model. I can change to {ID} | {name} if you suggest.

[erogol]

Yes, I know, but when selecting the 50th model, one needs to press down 50 times. But one could write 50 quickly :)

[p0p4k]

Although, the down button hard pressed, slides down the list real quick, I did not think it through!
Makes sense for longer lists to just type the ID.
Slightly busy with other things this week, will try to finish this within 2-3 weeks and request review. Thanks for the feedback, as usual! 🙌