Problem with TTS Larynx

I’m trying out Larynx under Rhasspy 2.5.10, but when I select “flemishguy” and vocoder quality “low quality”, then entering “test” and activating the speak button, I get the following error:

TtsException: [ONNXRuntimeError] : 3 : NO_SUCHFILE : Load model from /profiles/nl/tts/larynx/nl/flemishguy-glow_tts/generator.onnx failed:Load model /profiles/nl/tts/larynx/nl/flemishguy-glow_tts/generator.onnx failed. File doesn’t exist

Should I install something extra? How?

kind regards,
hugo

After saving your settings and restarting, make sure to look at the top of the web page for a “Download” button. This will download the right model files into your profile.

I wish I could’ve just included the models in the Docker image, but it was an extra few gigabytes :confused:

1 Like

OK, this solves it. I notice still some added t-sounds where they shouldn’t be e.g.
ik ga naar huis is pronounced as: ik ga naar huist
ik ga naar de bakker is pronounced as: ik ga naar de bakkert
jij moet opstaan is pronounced as: jij moet opstaant

Is there a chance response times will be reduced in the future?

kind regards,
hugo

1 Like

OK, thank you for the feedback. I’ve created a Github issue here: https://github.com/rhasspy/larynx/issues/15

A quick look suggests this is a problem with the voice models themselves. The pronunciation dictionary does not include a “t” and the ends of those words. Is this a regional thing?

1 Like

A quick look suggests this is a problem with the voice models themselves. The pronunciation dictionary does not include a “t” and the ends of those words. Is this a regional thing?

You shouldn’t hear a t there, no one will pronounce one there, neither people from Flanders nor from Holland
I also noticed when I use the tts for “wil” it adds the t, but in the short sentence “wil je”, wil is pronounced correctly.

And neither the voice of rdh nor bart_de_leeuw have this problem.

1 Like

Same error. In my case, if I try to download the entries it gives me the following error: DownloadFailedException: (‘https://github.com/rhasspy/gruut/releases/download/v0.9.0/it-it.tar.gz’, 'File size mismatch (got 4841389 byte (s), expected 4839488) ')
Rhasspy installed from .deb. Arch amd64.

Hi @gyppe,

I thought I had updated the .deb files, but maybe I missed one.

As a workaround, locate the file rhasspy-profile/rhasspyprofile/profiles/it/profile.json in the /usr/lib/rhasspy directory. Edit the file with sudo, and find fix this snippet at the bottom:

...
            "gruut/it-it.tar.gz": {
                "unzip": false,
                "url": "rhasspy/gruut/releases/download/v0.9.0/it-it.tar.gz",
                "bytes_expected": 4841389
            }

This is a general workaround for any “file system mismatch” error.

1 Like

Very thanks!!! :slight_smile: work perfect, and compliments for great software.

1 Like

You’re welcome :slight_smile:

1 Like

If you’re on Raspberry Pi, I just found out that by installing the 64 bit Raspberry Pi OS you can speed up Larynx quite significantly.