File size: 4,228 Bytes
489c533
 
 
 
 
60a3a10
489c533
 
 
 
 
 
 
 
 
 
60a3a10
 
489c533
 
 
 
60a3a10
489c533
 
 
 
 
 
60a3a10
 
 
 
 
a28c91e
60a3a10
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
489c533
60a3a10
 
489c533
 
 
 
 
 
60a3a10
489c533
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
# app.py - Gradio interface for Whisper Lao ASR
import gradio as gr
import torch
from transformers import WhisperProcessor, WhisperForConditionalGeneration
import numpy as np
import librosa

# Load model and processor
model_id = "Phonepadith/whisper-3-large-lao-finetuned-v1"
processor = WhisperProcessor.from_pretrained(model_id)
model = WhisperForConditionalGeneration.from_pretrained(model_id)

# Move to GPU if available
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)

print(f"Model loaded on: {device}")

def transcribe_audio(audio):
    """
    Transcribe audio to Lao text
    Args:
        audio: Audio file path (string) or tuple (sample_rate, audio_array) from Gradio
    Returns:
        transcription: Lao text
    """
    if audio is None:
        return "Please upload or record audio."
    
    try:
        # Handle both file paths and numpy arrays
        if isinstance(audio, str):
            # Audio is a file path - use librosa to load it
            audio_array, sample_rate = librosa.load(audio, sr=16000, mono=True)
        else:
            # Audio is a tuple (sample_rate, audio_array)
            sample_rate, audio_array = audio
            
            # Convert to float32 and normalize
            if audio_array.dtype != np.float32:
                # If integer type, normalize to [-1, 1]
                if np.issubdtype(audio_array.dtype, np.integer):
                    max_val = np.iinfo(audio_array.dtype).max
                    audio_array = audio_array.astype(np.float32) / max_val
                else:
                    audio_array = audio_array.astype(np.float32)
            
            # Ensure audio is in [-1, 1] range
            if np.abs(audio_array).max() > 1.0:
                audio_array = audio_array / np.abs(audio_array).max()
            
            # Resample to 16kHz if needed
            if sample_rate != 16000:
                audio_array = librosa.resample(
                    audio_array, 
                    orig_sr=sample_rate, 
                    target_sr=16000
                )
        
        # Process audio
        input_features = processor(
            audio_array,
            sampling_rate=16000,
            return_tensors="pt"
        ).input_features.to(device)
        
        # Generate transcription
        with torch.no_grad():
            predicted_ids = model.generate(input_features)
        
        # Decode transcription
        transcription = processor.batch_decode(
            predicted_ids, 
            skip_special_tokens=True
        )[0]
        
        return transcription
    
    except Exception as e:
        return f"Error processing audio: {str(e)}"

# Create Gradio interface
demo = gr.Interface(
    fn=transcribe_audio,
    inputs=gr.Audio(
        sources=["microphone", "upload"],
        type="filepath",  # Changed to filepath to handle various formats
        label="Record or Upload Lao Audio"
    ),
    outputs=gr.Textbox(
        label="Transcription (ພາສາລາວ)",
        placeholder="Your transcription will appear here...",
        lines=5
    ),
    title="🗣️ Whisper Large Lao ASR",
    description="""
    ### Automatic Speech Recognition for Lao Language
    
    This model transcribes Lao (ພາສາລາວ) speech to text using a fine-tuned Whisper Large model.
    
    **How to use:**
    1. Click the microphone icon to record audio, or upload an audio file
    2. Wait for the transcription to appear
    
    **Supported formats:** WAV, MP3, OGG, FLAC (16kHz recommended)
    """,
    article="""
    ### About this model
    
    - **Model:** Fine-tuned Whisper Large for Lao
    - **Dataset:** 7k+ Lao speech samples
    - **Repository:** [Phonepadith/whisper-3-large-lao-finetuned-v1](https://huggingface.co/Phonepadith/whisper-3-large-lao-finetuned-v1)
    
    ---
    
    Created by [@Phonepadith](https://huggingface.co/Phonepadith) | 📧 [email protected]
    """,
    examples=[
        # Add example audio files here if you have them
        # ["example1.wav"],
        # ["example2.wav"],
    ],
    cache_examples=False,
    theme=gr.themes.Soft()
)

# Launch the app
if __name__ == "__main__":
    demo.launch()