Kokoro TTS (Python) omits the word "specified"

[Deep0Thinking]

Hi! I've noticed that Kokoro TTS Python library consistently skips the word "specified" in the generated audio (I only tested for voice='am_michael' and voice='af_bella'), yet the online demo at hexgrad/Kokoro-TTS pronounces it correctly.

• Kokoro Version: 0.3.1
• OS: Windows

My python code:

import os
import numpy as np
import soundfile as sf
from kokoro import KPipeline
from IPython.display import display, Audio

with open("script.txt", "r", encoding="utf-8") as f:
    text = f.read()

pipeline = KPipeline(lang_code='a')

generator = pipeline(
    text, 
    voice='am_michael',
    speed=1,
    split_pattern=r'\n+'
)

os.makedirs('output', exist_ok=True)
combined_audio = []

for i, (gs, ps, audio) in enumerate(generator):
    print(i, gs)
    display(Audio(data=audio, rate=24000, autoplay=(i == 0)))

    combined_audio.append(audio)

full_audio = np.concatenate(combined_audio)
sf.write('output/full.wav', full_audio, 24000)

script.txt:

These are being specified in the text.

In the specified field.

Could you please help investigate why the word "specified" is dropped by the Python version, but not in the Hugging Face demo? Thanks!

[hexgrad]

@Deep0Thinking

1. Check if you have installed espeak-ng by running espeak-ng --version on your command line.
2. If you already had espeak-ng installed, follow RuntimeError: espeak not installed on your system bootphon/phonemizer#44 (comment) to make sure the espeak-ng path is known.
3. If you have done both of those things and specified is still being skipped, please also provide the print output of print(i, gs, ps) in your next troubleshooting comment.

[Deep0Thinking]

@hexgrad

It works, thanks a lot for the solution!

[UyCode]

@Deep0Thinking

1. Check if you have installed espeak-ng by running espeak-ng --version on your command line.
2. If you already had espeak-ng installed, follow RuntimeError: espeak not installed on your system bootphon/phonemizer#44 (comment) to make sure the espeak-ng path is known.
3. If you have done both of those things and specified is still being skipped, please also provide the print output of print(i, gs, ps) in your next troubleshooting comment.

I do installed the espeak-ng and it is in the PATH

this is my code, and result:

from kokoro import KPipeline
import soundfile as sf
import os

os.environ["PHONEMIZER_ESPEAK_LIBRARY"] = 'C:\Program Files\eSpeak NG\libespeak-ng.dll'
os.environ["_SEPEAK_LIBRARY"] = 'C:\Program Files\eSpeak NG\libespeak-ng.dll'
os.environ["PHONEMIZER_ESPEAK_PATH"] = 'C:\Program Files\eSpeak NG\espeak-ng.exe'

pipeline = KPipeline(lang_code='a')

text = '''
This is a specified Microsoft corporation with Kokoro
'''

generator = pipeline(
    text, voice='af_sarah', # <= change voice here
    speed=1, split_pattern=r'\n+'
)
for i, (gs, ps, audio) in enumerate(generator):
    print(i)  # i => index
    print(gs) # gs => graphemes/text
    print(ps) # ps => phonemes
    # display(Audio(data=audio, rate=24000, autoplay=i==0))
    sf.write(f'{i}.wav', audio, 24000) # save each audio file

result:
0
This is a specified Microsoft corporation with Kokoro
ðˌɪs ɪz ɐ kˌɔɹpəɹˈAʃən wɪð

[hexgrad]

@UyCode try the following:

from misaki import en, espeak
fallback = espeak.EspeakFallback(british=False)

[UyCode]

thank you for providing the help, it works