Update app.py
Browse files
app.py
CHANGED
@@ -1,8 +1,7 @@
|
|
1 |
import gradio as gr
|
|
|
2 |
import numpy as np
|
3 |
from pydub import AudioSegment
|
4 |
-
import io
|
5 |
-
from transformers import pipeline, AutoTokenizer
|
6 |
|
7 |
# Load the pipeline for speech recognition and translation
|
8 |
pipe = pipeline(
|
@@ -14,32 +13,22 @@ translator = pipeline("text2text-generation", model="Baghdad99/saad-hausa-text-t
|
|
14 |
tts = pipeline("text-to-speech", model="Baghdad99/english_voice_tts")
|
15 |
|
16 |
def translate_speech(audio_data_tuple):
|
|
|
|
|
17 |
# Extract the audio data from the tuple
|
18 |
sample_rate, audio_data = audio_data_tuple
|
19 |
|
20 |
-
#
|
21 |
-
|
22 |
-
|
23 |
-
# Create an AudioSegment from the audio data
|
24 |
-
audio_segment = AudioSegment(
|
25 |
-
audio_data_int16.tobytes(), # Audio data as bytes
|
26 |
-
frame_rate=sample_rate,
|
27 |
-
sample_width=audio_data_int16.dtype.itemsize, # Width in bytes
|
28 |
-
channels=1
|
29 |
-
)
|
30 |
|
31 |
-
#
|
32 |
-
|
33 |
-
audio_segment.export(mp3_buffer, format="mp3")
|
34 |
|
35 |
-
#
|
36 |
-
|
37 |
-
with open("audio.mp3", "wb") as f:
|
38 |
-
f.write(mp3_buffer.getvalue())
|
39 |
|
40 |
-
# Now you can feed the MP3 file to your model
|
41 |
# Use the speech recognition pipeline to transcribe the audio
|
42 |
-
output = pipe(
|
43 |
|
44 |
print(f"Output: {output}") # Print the output to see what it contains
|
45 |
|
@@ -91,6 +80,11 @@ def translate_speech(audio_data_tuple):
|
|
91 |
return 16000, synthesised_speech
|
92 |
|
93 |
|
|
|
|
|
|
|
|
|
|
|
94 |
# Define the Gradio interface
|
95 |
iface = gr.Interface(
|
96 |
fn=translate_speech,
|
|
|
1 |
import gradio as gr
|
2 |
+
from transformers import pipeline, AutoTokenizer
|
3 |
import numpy as np
|
4 |
from pydub import AudioSegment
|
|
|
|
|
5 |
|
6 |
# Load the pipeline for speech recognition and translation
|
7 |
pipe = pipeline(
|
|
|
13 |
tts = pipeline("text-to-speech", model="Baghdad99/english_voice_tts")
|
14 |
|
15 |
def translate_speech(audio_data_tuple):
|
16 |
+
print(f"Type of audio: {type(audio_data_tuple)}, Value of audio: {audio_data_tuple}") # Debug line
|
17 |
+
|
18 |
# Extract the audio data from the tuple
|
19 |
sample_rate, audio_data = audio_data_tuple
|
20 |
|
21 |
+
# Print the shape and type of the audio data
|
22 |
+
print(f"Audio data type: {type(audio_data)}, Audio data shape: {audio_data.shape}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
|
24 |
+
# Normalize the audio data to the range [-1, 1]
|
25 |
+
audio_data_normalized = audio_data / np.iinfo(audio_data.dtype).max
|
|
|
26 |
|
27 |
+
# Convert the normalized audio data to float64
|
28 |
+
audio_data_float64 = audio_data_normalized.astype(np.float64)
|
|
|
|
|
29 |
|
|
|
30 |
# Use the speech recognition pipeline to transcribe the audio
|
31 |
+
output = pipe(audio_data_float64)
|
32 |
|
33 |
print(f"Output: {output}") # Print the output to see what it contains
|
34 |
|
|
|
80 |
return 16000, synthesised_speech
|
81 |
|
82 |
|
83 |
+
# Define the Gradio interface
|
84 |
+
iface = gr.Interface(
|
85 |
+
fn=translate_speech,
|
86 |
+
inputs=gr.inputs.Audio(source="microphone"), # Change this line""
|
87 |
+
|
88 |
# Define the Gradio interface
|
89 |
iface = gr.Interface(
|
90 |
fn=translate_speech,
|