DocVoice.py

# DocVoice.py
import torch
from transformers import pipeline
import soundfile as sf

# -------------------
# 1️⃣ Detect GPU
# -------------------
use_cuda = torch.cuda.is_available()
device = 0 if use_cuda else -1
print(f"🌟 Using {'GPU' if use_cuda else 'CPU'}")

# -------------------
# 2️⃣ Load TTS model
# -------------------
tts_model_id = "microsoft/speecht5_tts"  # Compatible TTS model

tts_pipe = pipeline(
    "text-to-speech",
    model=tts_model_id,
    device=device
)

print("🔊 TTS pipeline ready using Hugging Face.")

# -------------------
# 3️⃣ TTS Helper Function
# -------------------
def text_to_speech(text: str, filename="assistant_response.wav"):
    """
    Generate speech from text and save as WAV file.
    """
    if not text.strip():
        return None

    print(f"📝 Generating audio for: {text}")

    speech_array = tts_pipe(text)[0]["array"]  # returns numpy array
    sample_rate = tts_pipe.model.config.sampling_rate if hasattr(tts_pipe.model.config, "sampling_rate") else 16000

    # Save audio
    sf.write(filename, speech_array, sample_rate)
    print(f"✅ Audio saved as {filename}")
    return filename