win 11 / linux - talk-llama voice interface is broken. Commit d6b9be2 from 19.01.2204

[mirek190]

Under windows 11 / linux no audio output
Commit d6b9be2 from 19.01.2204

.\talk-llama.exe -mw ggml-large-v3.bin -ml marcoroni-7b-v3.Q8_0.gguf -s speak.bat -ngl 99 -t 16

Variables $1 and $2 from speak / speak.bat are not passing proper strings
For instance I changed under windows speak.ps1 to

# Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope CurrentUser
param(
  # voice options are David or Zira
  [Parameter(Mandatory=$true)][string]$voice,
  [Parameter(Mandatory=$true)][string]$text
)

# testing variables
Add-Content -Path C:\AI_music\output.txt -Value $voice
Add-Content -Path C:\AI_music\output.txt -Value $text

Add-Type -AssemblyName System.Speech;
$speak = New-Object System.Speech.Synthesis.SpeechSynthesizer;
$speak.SelectVoice("Microsoft $voice Desktop");
$speak.Rate="0";
$speak.Speak($text);

to get variables $1 and $2 to a file

Output the output.txt during talking with talk-llama

2
'
2
'
2
'
2
'
2
'

Conclusion - audio interface is broken.

[mirek190]

Hello

any chance to fix it?

[tamo]

Seems that speak.bat expects only two arguments but talk-llama gives many.

1. You can edit speak.bat to pass through all the args (@powershell -ExecutionPolicy Bypass -F examples\talk\speak.ps1 %*) and edit speak.ps1 like

...
param(
  # voice options are David or Zira
  [Parameter(Mandatory=$true, Position=0)][string]$voice,
  [Parameter(Mandatory=$true, Position=1, ValueFromRemainingArguments=$true)][string]$text
)
...

2. Or you can edit speak.bat like

@echo off
for /f "tokens=1*" %%a in ("%*") do (
  set voice=%%a
  set text=%%b
)
powershell -ExecutionPolicy Bypass -F examples\talk\speak.ps1 %voice% "%text%"

I found other problems (they should be handled as other issues though)

1. voice should be "David or Zira" but talk-llama specifies just "2"

Exception calling "SelectVoice" with "1" argument(s): "Cannot set voice. No matching voice is installed or the voice wa
s disabled."
At speak.ps1:10 char:1
+ $speak.SelectVoice("Microsoft $voice Desktop");
+ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    + CategoryInfo          : NotSpecified: (:) [], MethodInvocationException
    + FullyQualifiedErrorId : ArgumentException

2. talk-llama should delete newlines in text_to_speak before passing it to system() (otherwise speak.bat stops speaking after the first line)

[ggerganov]

I don't have a Windows machine, so hoping someone else to help with this

[mirek190]

Under Linux audio also not working with talk-llama

[tamo]

Maybe text_to_speak should be passed as file
https://github.com/ggerganov/whisper.cpp/compare/master...tamo:whisper.cpp:passfile?expand=1
I have tested it only with speak.bat though