File size: 2,800 Bytes
6c20eaa
38b530f
 
 
 
3ac0359
6c20eaa
38b530f
 
 
6c20eaa
38b530f
 
6c20eaa
38b530f
 
 
 
 
 
 
 
6c20eaa
38b530f
 
 
 
809a47f
38b530f
6c20eaa
38b530f
 
6c20eaa
809a47f
26107f3
 
809a47f
26107f3
 
809a47f
26107f3
 
6c20eaa
3ac0359
 
 
 
 
809a47f
 
 
 
 
3ac0359
 
809a47f
3ac0359
 
 
 
809a47f
38b530f
 
 
3ac0359
38b530f
 
6c20eaa
38b530f
 
6c20eaa
809a47f
 
 
 
3ac0359
809a47f
6c20eaa
3ac0359
38b530f
 
 
 
 
809a47f
38b530f
 
 
 
6c20eaa
38b530f
 
 
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
import gradio as gr
import tensorflow as tf
import numpy as np
import nltk
from nltk.corpus import cmudict
from scipy.io.wavfile import write

# Download required NLTK data
nltk.download('averaged_perceptron_tagger')
nltk.download('cmudict')

# Load your model
model = tf.keras.models.load_model('audio_model.h5')

# Preprocess input text
def preprocess_text(text):
    """
    Process the input text to prepare it for the model.
    """
    d = cmudict.dict()
    words = text.lower().split()
    phonemes = []

    for word in words:
        if word in d:
            phonemes.append(d[word][0])
        else:
            # Use a placeholder for words not found in cmudict
            phonemes.append(['UNKNOWN'])
    
    # Flatten the list of phonemes
    flattened_phonemes = [p for sublist in phonemes for p in sublist]
    
    # Create dummy feature vectors (this should be replaced with actual feature extraction)
    num_features = 13
    sequence_length = len(flattened_phonemes)
    input_data = np.random.rand(sequence_length, num_features)  # Placeholder
    
    # Add batch dimension
    input_data = np.expand_dims(input_data, axis=0)
    
    return input_data

# Convert model output to an audio file
def convert_to_audio(model_output, filename="output.wav", sample_rate=22050):
    """
    Convert the model output into a .wav file.
    """
    # Check if model_output is empty
    if model_output is None or len(model_output) == 0:
        raise ValueError("Model output is empty.")

    # Normalize the audio output
    normalized_output = np.interp(model_output, (model_output.min(), model_output.max()), (-1, 1))

    # Write the audio data to a file
    write(filename, sample_rate, normalized_output)

    return filename

# Generate sound effect
def generate_sfx(text):
    """
    Takes input text, preprocesses it, runs it through the model,
    and generates a downloadable audio file.
    """
    input_data = preprocess_text(text)
    
    # Generate prediction
    prediction = model.predict(input_data)

    # Ensure prediction shape is correct
    if prediction.ndim == 2 and prediction.shape[1] > 1:
        prediction = prediction.flatten()  # Flatten if necessary

    # Convert the prediction to an audio file
    audio_file = convert_to_audio(prediction, filename="output.wav")
    
    return audio_file

# Define the Gradio interface
interface = gr.Interface(
    fn=generate_sfx,
    inputs=gr.Textbox(label="Enter a Word", placeholder="Write a Word To Convert it into SFX Sound"),
    outputs=gr.Audio(label="Generated SFX", type="filepath"),
    live=False,
    title="SFX Generator from Text",
    description="Enter a word or sentence, and the model will generate an SFX sound.",
)

# Run the interface
if __name__ == "__main__":
    interface.launch()