[Python] can't generate the speech from library tts of coqui-tts and this error happens in...

from TTS.utils.manage import ModelManager
from TTS.utils.synthesizer import Synthesizer

from google.colab import files

Initialize the model manager and load the models

model_name = "tts_models/en/ljspeech/tacotron2-DDC"

vocoder_name = "vocoder_models/en/ljspeech/hifigan_v2"

model_manager = ModelManager()

model_path, config_path, _ = model_manager.download_model(model_name)

vocoder_path, vocoder_config_path, _ = model_manager.download_model(vocoder_name)

Create the synthesizer object

synthesizer = Synthesizer(model_path, config_path, vocoder_path, vocoder_config_path, use_cuda=False)

Generate dynamic SSML based on manually chosen emotion

def generate_dynamic_ssml(chunk):

ssml = f"""<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xml:lang='en-US'>"""



# Uncomment the emotion you want to apply

ssml += f"<prosody pitch='+10%' rate='fast'><emphasis level='moderate'>{chunk}</emphasis></prosody>" # happy

# ssml += f"<prosody pitch='+5%' rate='medium'><emphasis level='strong'>{chunk}</emphasis></prosody>" # romantic

# ssml += f"<prosody pitch='+7%' rate='medium'><emphasis level='moderate'>{chunk}</emphasis></prosody>" # hopeful

# ssml += f"<prosody pitch='default' rate='medium'><emphasis level='none'>{chunk}</emphasis></prosody>" # neutral

# ssml += f"<prosody pitch='-5%' rate='slow'><emphasis level='moderate'>{chunk}</emphasis></prosody>" # disappointed

# ssml += f"<prosody pitch='-10%' rate='fast'><emphasis level='strong'>{chunk}</emphasis></prosody>" # angry

# ssml += f"<prosody pitch='-10%' rate='slow'><emphasis level='strong'>{chunk}</emphasis></prosody>" # fearful

# ssml += f"<prosody pitch='-15%' rate='slow'><emphasis level='strong'>{chunk}</emphasis></prosody>" # sad

# ssml += f"<prosody pitch='-20%' rate='very slow'><emphasis level='strong'>{chunk}</emphasis></prosody>" # devastated



ssml += "</speak>"

return ssml

Function to synthesize speech for each chunk

def synthesize_speech(text):

# Generate SSML for the whole text with a manually selected emotion

ssml = generate_dynamic_ssml(text)



# Synthesize the text

wav = synthesizer.tts(ssml)



# Save the output file

output_file = "output_with_emotion.wav"

synthesizer.save_wav(wav, output_file)



# Download the file to the local machine

files.download(output_file)

Example usage

if name == "main":

sample_text = """I am so happy to see you! You make my heart race with joy and love."""



# Convert text to speech with the selected emotion

synthesize_speech(sample_text)


code using tacotron2 model and vocoder hifigen after that I use the synthesizer library ssml that change the pitch of the voice to make it like emotion

AttributeError Traceback (most recent call last)

<ipython-input-8-2698293d91f3> in <cell line: 51>()

53

54 # Convert text to speech with the selected emotion

---> 55 synthesize_speech(sample_text)

1 frames

<ipython-input-8-2698293d91f3> in synthesize_speech(text)

39

40 # Synthesize the text

---> 41 wav = synthesizer.tts(ssml)

42

43 # Save the output file

/usr/local/lib/python3.10/dist-packages/TTS/utils/synthesizer.py in tts(self, text, speaker_name, language_name, speaker_wav, style_wav, style_text, reference_wav, reference_speaker_name, split_sentences, **kwargs)

320 speaker_id = self.tts_model.speaker_manager.name_to_id[speaker_name]

321 # handle Neon models with single speaker.

--> 322 elif len(self.tts_model.speaker_manager.name_to_id) == 1:

323 speaker_id = list(self.tts_model.speaker_manager.name_to_id.values())[0]

324 elif not speaker_name and not speaker_wav:

AttributeError: 'NoneType' object has no attribute 'name_to_id' that is the error



that tts code using coqui-tts that make emotional tts so the following error always apper and I have tried to make the code single and multi speakers

make single and multi speakers maybe someone know the solution or advice me to use another model I want it a free one

Continue reading...