Replies: 8 comments 20 replies
-
|
Is it possible to prevent piper from terminating and restarting after every sentence, it takes a lot of time after each sentence? I saw this in the log of speech-dispatcher. |
Beta Was this translation helpful? Give feedback.
-
|
Hey @takov751 this works pretty well for being "hacked together". I've just encountered two issues so far. The first is that the audio stops if I write a long sentence using spd-say, so it doesn't finish saying it. And the other is that apps that connect to speech-dispatcher like firefox and foliate give me an error saying they can't use speech-dispatcher. Does anyone know how these issues can be resolved? |
Beta Was this translation helpful? Give feedback.
-
|
Nevermind, after a few reboots it's working fine |
Beta Was this translation helpful? Give feedback.
-
|
That is interesting, thanks for reporting back. I forgot to add different language models so that it can be addressed by other apps even if it's hardcoded |
Beta Was this translation helpful? Give feedback.
-
|
Could you please give some description to non-technical users like me: what these values are and how can I replace them to chose different voice? How to find this classification for equivalent of "MALE1", for example I typed For example in Plasma Okular there is an option to change voice (this also sometimes means language). But with proposed configuration I don't know how to make other voices available for speech dispatcher. |
Beta Was this translation helpful? Give feedback.
-
|
You are making a very good point. my example was hard coded and did not calculated this in. I have modified my first post so that it accommodates those specific voice options. this way you are able to add even different languages as well. and you will be able to choose between the voices with just |
Beta Was this translation helpful? Give feedback.
-
|
I want to use another voice(german) instead the default one(english): not sure why I still hear the default english voice. My code should work and use the german voice I've installed. If I change the default voice under pied's settings to german, all english articles sound wrong, of course. (OS:Mint). |
Beta Was this translation helpful? Give feedback.
-
|
Could you show us your /etc/speech-dispatcher/modules/piper.conf ? |
Beta Was this translation helpful? Give feedback.
-
|
Yes, sure! |
Beta Was this translation helpful? Give feedback.
-
So i think in your case the issue is that VOICE parameter is hardcoded. Your file should look like this This way you will have optional voice model |
Beta Was this translation helpful? Give feedback.
-
|
You only showed piper.conf. Check speechd.conf. In the link I provided above you have examples. My contains: And the command you provided works for me for both languages. Also in E-book viewer I have all voices available but in Dolphin I can have only English. But if you use Pied just move all configurations elsewhere and Pied should create both config files. It will recreate them every time you change Pied settings. |
Beta Was this translation helpful? Give feedback.
-
I did not include the contents of
I'm glad they worked for you
yes, I thought the same, best will be to save both configurations somewhere and rewrite the main I tried both Takov's and your configurations correcting my user path, but the terminal uses always the same default voice set in Pied. |
Beta Was this translation helpful? Give feedback.
-
|
Hi, I followed this guide (I think) but it doesn't work. When I try this command (just for testing), it works. But the integration with speechd doesn't. Here is my config
In the logs I can see
It seems that piper cannot find the model's config even though it is at the right place. |
Beta Was this translation helpful? Give feedback.
-
Interesting. I have no idea at this point. Right now i have very similar setup and mine works when i specify language( for some reason it does not work without it) like |
Beta Was this translation helpful? Give feedback.
-
|
I'm on NixOS 24.05 with Pipewire enabled. If you are interested in it and you know Nix, my config is public (https://codeberg.org/Cameo007/NixOS-Configs) |
Beta Was this translation helpful? Give feedback.
-
I am familiar with nix and i have a suspicion that it could be a speech-dispatcher packaging issue or at least some sort of access issue. |
Beta Was this translation helpful? Give feedback.
-
|
Nice |
Beta Was this translation helpful? Give feedback.
-
|
@Cameo007 I had the same issue. with |
Beta Was this translation helpful? Give feedback.
-
|
On arch installing and rebooting was enough for it to work. It creates a which for some reason isn't explicitly loaded into the |
Beta Was this translation helpful? Give feedback.
-
|
i have a blind user running ubuntu (and well actually i cannot install arch there) and wonder what else the package contains. Is there f.e. a place for the models that piper will always find? rn i use a subpath of /etc/speech-dispatcher |
Beta Was this translation helpful? Give feedback.
-
More info here In addition, it also packages a script, |
Beta Was this translation helpful? Give feedback.
-
I believe that if there aren't any P.S. I haven't tested this, nor could I find it in the |
Beta Was this translation helpful? Give feedback.
-
|
I think I finally made it work in nixOS @Cameo007 {
config,
lib,
pkgs,
...
}:
let
cfg = config.mySystem.piperTextToSpeech;
user = "yeshey";
rmNewLines = pkgs.writeShellScriptBin "rmNewLines"
''
echo "''${1//$'\n'/ }"
'';
piperVoices = {
en_US_amy = {
onnx = builtins.fetchurl {
url = "https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/amy/medium/en_US-amy-medium.onnx?download=true";
sha256 = "sha256:063c43bbs0nb09f86l4avnf9mxah38b1h9ffl3kgpixqaxxy99mk"; # Replace with the actual sha256 hash after running `nix-build`.
};
json = builtins.fetchurl {
url = "https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/amy/medium/en_US-amy-medium.onnx.json";
sha256 = "sha256:0xvxjxk59byydx9gj6rdvvydp5zm8mzsrf9vyy6x6299sjs3x8lm"; # Replace with the actual sha256 hash after running `nix-build`.
};
};
en_US_libritts_r = {
onnx = builtins.fetchurl {
url = "https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/libritts_r/medium/en_US-libritts_r-medium.onnx?download=true";
sha256 = "sha256:159iq7x4idczq4p5ap9wmf918jfhk4brydhz0zsgq5nnf7h8bfqh"; # Replace with the actual sha256 hash after running `nix-build`.
};
json = builtins.fetchurl {
url = "https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/libritts_r/medium/en_US-libritts_r-medium.onnx.json";
sha256 = "sha256:1cxgr5dm0y4q4rxjal80yhbjhydzdxnijg9rkj0mwcyqs9hdqwdl"; # Replace with the actual sha256 hash after running `nix-build`.
};
};
};
createFile = { name, content }: ''
mkdir -p "$(dirname /home/${user}/.config/${name})"
echo '${content}' > /home/${user}/.config/${name}
'';
in
{
options.mySystem.piperTextToSpeech = {
enable = lib.mkEnableOption "Enable Piper Text-to-Speech.";
};
config = lib.mkIf (config.mySystem.enable && cfg.enable) {
# Use userActivationScripts to configure Piper and Speech Dispatcher
system.userActivationScripts.piperSpeechDispatcher = ''
# Create Piper configuration in user-specific locations
${createFile {
name = "speech-dispatcher/modules/piper.conf";
content = ''
GenericExecuteSynth '\''''''echo "$(${rmNewLines}/bin/rmNewLines "$DATA")" | ${pkgs.piper-tts}/bin/piper --model /etc/piper-voices/en_US-libritts_r-medium.onnx --output_raw --sentence_silence 0.1 | ${pkgs.sox}/bin/play -t raw -r 22050 -e signed-integer -b 16 -c 1 - tempo $(echo "1 + ($RATE / 100) * (2.5 - 1)" | ${pkgs.bc}/bin/bc -l) '\''''''
AddVoice "en-US" "male1" "en/en_US/libritts_r/medium/en_US-libritts_r-medium.onnx"
'';
}}
${createFile {
name = "speech-dispatcher/speechd.conf";
content = ''
AddModule "piper" "sd_generic" "piper.conf"
DefaultVoiceType "male1"
DefaultLanguage en-US
DefaultModule piper
'';
}}
'';
#environment.etc."piper-voices/en_US-amy-medium.onnx".source = piperVoices.en_US_amy.onnx;
#environment.etc."piper-voices/en_US-amy-medium.onnx.json".source = piperVoices.en_US_amy.json;
environment.etc."piper-voices/en_US-libritts_r-medium.onnx".source = piperVoices.en_US_libritts_r.onnx;
environment.etc."piper-voices/en_US-libritts_r-medium.onnx.json".source = piperVoices.en_US_libritts_r.json;
# Ensure Piper and Speech Dispatcher are installed
environment.systemPackages = with pkgs; [
speechd
piper-tts
pipewire
sox
bc
];
};
}Edit: Removing new lines so it doesnt mess up reading pdfs |
Beta Was this translation helpful? Give feedback.
-
|
I have found one major issue in using Does anyone have a solution for this? perhaps a process that runs in the background that loads the model and then simply waits for input? |
Beta Was this translation helpful? Give feedback.
-
|
Best I did was load in the next phrase as the previous phrase was still speaking, but it became unable to cancel ongoing utterances. |
Beta Was this translation helpful? Give feedback.
-
I was able to hack it together, it was rather simple.
Install speech-dispatcher then in
/etc/speech-dispatcher/modules/piper.conf/etc/speech-dispatcher/speechd.confand you have piper as narrator on your desktop. Of course there is lot of room for improvement, but slowly getting there
/path/to/voice/should be the directory you have downloaded the voices from huggingfaceBeta Was this translation helpful? Give feedback.
All reactions