cannot import name 'Coqpit' from 'coqpit' (unknown location)

[user3691]

I am struggling with the Coqui TTS, because the first time I installed the original TTS ran into errors. Then I found out that the original is not maintained anymore, therefore installed this version. Now when trying to run:

joe@joe-pc:~/coqui-ai-TTS$ tts --list_models
Traceback (most recent call last):
File "/home/joe/.local/bin/tts", line 5, in
from TTS.bin.synthesize import main
File "/home/joe/coqui-ai-TTS/TTS/init.py", line 24, in
from TTS.config.shared_configs import BaseDatasetConfig
File "/home/joe/coqui-ai-TTS/TTS/config/init.py", line 8, in
from coqpit import Coqpit
ImportError: cannot import name 'Coqpit' from 'coqpit' (unknown location)

Get an error again. What is wrong?

[eginhard]

Duplicate of #300. You can try the steps mentioned there, i.e. purging the pip cache and recreating your virtual environment.

[eginhard]

You could try with python -m pip install coqpit-config instead to make sure it uses the correct Python. If it still doesn't work I'd recommend installing coqui-tts into a virtual environment to ensure everything is properly isolated.

[user3691]

Here is what happened:

joe@joe-pc:/coqui-ai-TTS$ python -m pip install coqpit-config
Command 'python' not found, did you mean:
command 'python3' from deb python3
command 'python' from deb python-is-python3
joe@joe-pc:/coqui-ai-TTS$ python3 -m pip install coqpit-config
Defaulting to user installation because normal site-packages is not writeable
Requirement already satisfied: coqpit-config in /home/joe/.local/lib/python3.11/site-packages (0.2.0)
Requirement already satisfied: typing-extensions>=4.10 in /home/joe/.local/lib/python3.11/site-packages (from coqpit-config) (4.12.2)

[eginhard]

Use a virtual environment then to install dependencies. If you're not sure what this means, the easiest way would be to install uv and do uv run tts --list_models.

[user3691]

I have uninstalled tts and deleted all folders containing this name. Then purged the catch.
Next installed uv and created an environment, activated it, then installed the coqui-tts with the line:
uv pip install coqui-tts

Now I was able to get a list of models with the instruction:
(joe) joe@joe-pc:~$ tts --list_models

The next step would be test the basic TTS conversion. I have tried this:
(joe) joe@joe-pc:~$ tts --text "Hello! How are you my friend?" --out_path home/joe/MY_TTS/testSpeech_001.wav

It started working on just fine, but after a while stopped with this:

Generator model: hifigan_generator
Discriminator model: hifigan_discriminator
Removing weight norm...
Text: THello! How are you my friend?
Text split into sentences.
Input: ['THello!', 'How are you my friend?']
[W301 16:40:46.877411789 NNPACK.cpp:62] Could not initialize NNPACK! Reason: Unsupported hardware.
Processing time: 12.975████████████████████| 3.80M/3.80M [00:16<00:00, 1.63MiB/s]
Real-time factor: 4.107
Traceback (most recent call last):
File "/home/joe/.venv/bin/tts", line 10, in
sys.exit(main())
^^^^^^
File "/home/joe/.venv/lib/python3.11/site-packages/TTS/bin/synthesize.py", line 413, in main
api.tts_to_file(
File "/home/joe/.venv/lib/python3.11/site-packages/TTS/api.py", line 385, in tts_to_file
self.synthesizer.save_wav(wav=wav, path=file_path, pipe_out=pipe_out)
File "/home/joe/.venv/lib/python3.11/site-packages/TTS/utils/synthesizer.py", line 275, in save_wav
save_wav(wav=wav, path=path, sample_rate=self.output_sample_rate, pipe_out=pipe_out)
File "/home/joe/.venv/lib/python3.11/site-packages/TTS/utils/audio/numpy_transforms.py", line 471, in save_wav
scipy.io.wavfile.write(path, sample_rate, wav_norm)
File "/home/joe/.venv/lib/python3.11/site-packages/scipy/io/wavfile.py", line 793, in write
fid = open(filename, 'wb')
^^^^^^^^^^^^^^^^^^^^
FileNotFoundError: [Errno 2] No such file or directory: 'home/joe/MY_TTS/testSpeech_001.wav'

My laptop does not have any GPU except for the one built into the processor. But, I was expecting that it should work using CPU only if it does not find any GPU. At lest I was able to run this part successfully using the original coqui-tts. What can I do to get it working?

---

System:
Kernel: 6.1.0-16-amd64 arch: x86_64 bits: 64 compiler: gcc v: 12.2.0 Desktop: Cinnamon v: 5.8.4
tk: GTK v: 3.24.38 wm: muffin dm: LightDM Distro: LMDE 6 Faye base: Debian 12.1 bookworm
Machine:
Type: Laptop System: LENOVO product: 80TG v: Lenovo V110-15IAP serial:
Chassis: type: 10 v: Lenovo V110-15IAP serial:
Mobo: LENOVO model: LNVNB161216 v: NO DPK serial: UEFI: LENOVO v: 1MCN42WW
date: 10/26/2017
Battery:
ID-1: BAT0 charge: 24.0 Wh (100.0%) condition: 24.0/24.0 Wh (99.9%) volts: 12.3 min: 10.8
model: LGC L15L3A02 serial: status: full
CPU:
Info: dual core model: Intel Celeron N3350 bits: 64 type: MCP arch: Goldmont rev: 9 cache:
L1: 112 KiB L2: 2 MiB
Speed (MHz): avg: 2307 high: 2326 min/max: 800/2400 cores: 1: 2288 2: 2326 bogomips: 4377
Flags: ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx
Graphics:
Device-1: Intel HD Graphics 500 vendor: Lenovo driver: i915 v: kernel arch: Gen-9 ports:
active: HDMI-A-1 off: eDP-1 empty: none bus-ID: 00:02.0 chip-ID: 8086:5a85
Display: x11 server: X.Org v: 1.21.1.7 with: Xwayland v: 22.1.9 driver: X: loaded: modesetting
unloaded: fbdev,vesa dri: iris gpu: i915 display-ID: :0 screens: 1
Screen-1: 0 s-res: 1280x1024 s-dpi: 96
Monitor-1: HDMI-A-1 mapped: HDMI-1 pos: primary model: Fujitsu Siemens 43B1-M res: 1280x1024
dpi: 96 diag: 434mm (17.1")
Monitor-2: eDP-1 note: disabled model: AU Optronics 0x71ec res: 1280x1024 dpi: 101
diag: 394mm (15.5")
API: OpenGL v: 4.6 Mesa 22.3.6 renderer: Mesa Intel HD Graphics 500 (APL 2) direct-render: Yes
Audio:
Device-1: Intel Celeron N3350/Pentium N4200/Atom E3900 Series Audio Cluster vendor: Lenovo
driver: snd_hda_intel v: kernel bus-ID: 00:0e.0 chip-ID: 8086:5a98
API: ALSA v: k6.1.0-16-amd64 status: kernel-api
Server-1: PipeWire v: 0.3.65 status: active with: 1: pipewire-pulse status: active
2: wireplumber status: active 3: pipewire-alsa type: plugin

[eginhard]

Per the error message No such file or directory: 'home/joe/MY_TTS/testSpeech_001.wav' the folder home/joe/MY_TTS/ doesn't exist. Probably it's a typo and you meant /home/joe/MY_TTS/? Otherwise create it first.

[user3691]

Well, it shows a file not found error, which makes no sense because that is the output file name that the tts supposed to create.
The folder home/joe/MY_TTS does exist and in that folder I am expecting the program to create teh output file named testSpeech_001.wav

But the first error occurs earlier, namely here:
[W301 16:40:46.877411789 NNPACK.cpp:62] Could not initialize NNPACK! Reason: Unsupported hardware.

This is what breaks the processing. I think I will give up on this package, because there is no end to errors using this software on my laptop. Will look for something else that works out of the box.

[eginhard]

The NNPACK message is just a warning, as you can see from the following lines the processing actually proceeds correctly. It fails because the folder home/joe/MY_TTS/ doesn't exist, please try again with the slash in front of home like this: tts --text "Hello! How are you my friend?" --out_path /home/joe/MY_TTS/testSpeech_001.wav

[user3691]

That worked, thanks. Finally I was able to generate the first speech output. Originally I added the first slash, but when it reported an error I thought I will try removing it, and then it remained that way in my later experiments.

Although the default output is of very poor quality. See the attached files. The testSpeech_000.wav was produced outside of uv environment, the testSpeech_001.wav was done within uv environment.

Anyway, I will experiment now with the other models and see if there is anything of good quality. Originally I tested this demo and was impressed with the performance:
https://huggingface.co/spaces/coqui/xtts

That is why I wanted to install tts locally. Will check if I can get that working now.
Thanks for your help, I highly appreciate it.

sample-files.zip

[eginhard]

Yes, the default model (tts_models/en/ljspeech/tacotron2-DDC) isn't the best, it's just to be able to test it easily. To run XTTS also see https://coqui-tts.readthedocs.io/en/latest/models/xtts.html