PaitentVoiceToText.py

# PaitentVoiceToText.py
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline

# -------------------
# 1️⃣ Detect device
# -------------------
use_cuda = torch.cuda.is_available()
dtype = torch.float16 if use_cuda else torch.float32

print(f"🌟 Using {'GPU' if use_cuda else 'CPU'}, dtype={dtype}")

# -------------------
# 2️⃣ Load Whisper model
# -------------------
hub_id = "Muhammadidrees/WispherVOICE"

print("⏳ Loading model...")
model = AutoModelForSpeechSeq2Seq.from_pretrained(
    hub_id,
    torch_dtype=dtype,
    device_map="auto",         # accelerate manages device placement
    trust_remote_code=True
)

processor = AutoProcessor.from_pretrained(
    hub_id,
    trust_remote_code=True
)

# -------------------
# 3️⃣ Create pipeline (no device argument!)
# -------------------
pipe = pipeline(
    "automatic-speech-recognition",
    model=model,
    tokenizer=processor.tokenizer,
    feature_extractor=processor.feature_extractor
)

print("🎧 Whisper pipeline ready.")

# -------------------
# 4️⃣ Function for external import
# -------------------
def record_and_transcribe(audio_file):
    """
    Transcribe an audio file (path) or recording.
    Returns the transcribed text.
    """
    if audio_file is None:
        return "No audio provided."
    result = pipe(audio_file)
    return result["text"]