AllTalk_TTS giving errors when run using ./start_alltalk.sh in ubuntu 22.04

[ahmetax]

Hi. I installed standalone AllTalk_TTS on my Ubuntu 22.04. It gave no errors during installation. However when I run the shell script start_alltalk.sh, I get lots of errors.
I starts as follows:
[AllTalk Startup] _ _ _ _____ _ _ _____ _____ ____
[AllTalk Startup] / \ | | |_ | | | | __ | | / |
[AllTalk Startup] / _ \ | | | | |/ ` | | |/ / | | | | _ \
[AllTalk Startup] / ___ | | | | | (| | | < | | | | ) |
[AllTalk Startup] // __|| ||_,|||_\ || || |____/
[AllTalk Startup]
[AllTalk Startup] Config file check : No Updates required
Traceback (most recent call last):
File "/home/axax/github/alltalk_tts/script.py", line 118, in
from TTS.api import TTS
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/TTS/api.py", line 9, in
from TTS.utils.audio.numpy_transforms import save_wav
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/TTS/utils/audio/init.py", line 1, in
from TTS.utils.audio.processor import AudioProcessor
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/TTS/utils/audio/processor.py", line 10, in
from TTS.utils.audio.numpy_transforms import (
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/TTS/utils/audio/numpy_transforms.py", line 8, in
from librosa import magphase, pyin
File "", line 1229, in _handle_fromlist
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/lazy_loader/init.py", line 78, in getattr
attr = getattr(submod, name)
^^^^^^^^^^^^^^^^^^^^^
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/lazy_loader/init.py", line 77, in getattr
submod = importlib.import_module(submod_path)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/importlib/init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/librosa/core/spectrum.py", line 16, in
from .audio import resample
File "/home/axax/github/alltalk_tts/alltalk_environment/env/lib/python3.11/site-packages/librosa/core/audio.py", line 1152, in
@guvectorize(
^^^^^^^^^^^^
.........
At the last line of the errors it says: "EOFError: Ran out of input"
What should I do? Thanks.

[erew123]

Hi @ahmetax

That's certainly an unusual one Ive not seen before. Im guessing you are using V1 of AllTalk v1 and not the AllTalk v2 BETA, as the V1 does an import of TTS on line 118. Both should work fine though (I both use and tested on Ubuntu 22.04).

I can think of 2x things to try/check.

1. Run ./start_environment.sh and then run pip install librosa==0.10.1

You can also try:

2. Run ./start_environment.sh and then run pip uninstall tts and then pip install coqui-tts

After either of these you should be able to do pip show librosa or pip show coqui-tts and they should both come back showing they are installed and librosa should should its version 0.10.1.

At this moment in time, Im thinking something has happened with some updates/changes. I can see librosa have made an update just the other month and maybe that update is doing something funny.

Beyond that, you would have to send me a diagnostics log so that I could look through it and try identify what could be going on OR you could try the BETA version of AllTalk (ensure you follow the Linux instructions if you do)..

I am going to be travelling not long after sending you this message, so my availability to look at anything/respond will be somewhat limited today/tomorrow, but feel free to get back to me and I will respond when I can.

Thanks

[ahmetax]

Thanks erew123.
I applied both of your suggestions, but, my problem continues.
Now I will try BETA version.
I wish you a good travel.

[ahmetax]

Hi erew123,
As you suggested, I installed an tried the BETA version, and everything is perfect now!
I tried some Turkish text, and the result is the best local solution.
Thanks a lot!