Spaces:
Sleeping
Sleeping
Jaward
commited on
Commit
·
535db75
1
Parent(s):
0e30afa
app
Browse files- app.py +65 -0
- requirements.txt +7 -0
app.py
ADDED
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import subprocess
|
3 |
+
import gradio as gr
|
4 |
+
|
5 |
+
# Supported languages
|
6 |
+
LANGUAGE_CODES = {
|
7 |
+
"English": "eng",
|
8 |
+
"Spanish": "spa",
|
9 |
+
"French": "fra",
|
10 |
+
"German": "deu",
|
11 |
+
"Italian": "ita",
|
12 |
+
"Chinese": "cmn"
|
13 |
+
}
|
14 |
+
|
15 |
+
def translate_speech(audio_file, target_language):
|
16 |
+
"""
|
17 |
+
Translate input speech (audio file) to the specified target language.
|
18 |
+
|
19 |
+
Args:
|
20 |
+
audio_file (str): Path to the input audio file.
|
21 |
+
target_language (str): The target language for translation.
|
22 |
+
|
23 |
+
Returns:
|
24 |
+
str: Path to the translated audio file.
|
25 |
+
"""
|
26 |
+
language_code = LANGUAGE_CODES[target_language]
|
27 |
+
output_file = "translated_audio.wav"
|
28 |
+
|
29 |
+
command = [
|
30 |
+
"expressivity_predict",
|
31 |
+
audio_file,
|
32 |
+
"--tgt_lang", language_code,
|
33 |
+
"--model_name", "seamless_expressivity",
|
34 |
+
"--vocoder_name", "vocoder_pretssel",
|
35 |
+
"--gated-model-dir", "seamlessmodel",
|
36 |
+
"--output_path", output_file
|
37 |
+
]
|
38 |
+
|
39 |
+
subprocess.run(command, check=True)
|
40 |
+
|
41 |
+
if os.path.exists(output_file):
|
42 |
+
print(f"File created successfully: {output_file}")
|
43 |
+
else:
|
44 |
+
print(f"File not found: {output_file}")
|
45 |
+
|
46 |
+
return output_file
|
47 |
+
|
48 |
+
def create_interface():
|
49 |
+
"""Create and configure the Gradio interface."""
|
50 |
+
inputs = [
|
51 |
+
gr.Audio(type="filepath", label="Audio File"),
|
52 |
+
gr.Dropdown(list(LANGUAGE_CODES.keys()), label="Target Language")
|
53 |
+
]
|
54 |
+
|
55 |
+
return gr.Interface(
|
56 |
+
fn=translate_speech,
|
57 |
+
inputs=inputs,
|
58 |
+
outputs=gr.Audio(label="Translated Audio"),
|
59 |
+
title="Seamless Expressive Speech-To-Speech Translator",
|
60 |
+
description="Hear how you sound in another language.",
|
61 |
+
)
|
62 |
+
|
63 |
+
if __name__ == "__main__":
|
64 |
+
iface = create_interface()
|
65 |
+
iface.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
git+https://github.com/facebookresearch/seamless_communication.git
|
2 |
+
pydub
|
3 |
+
soundfile
|
4 |
+
torchaudio
|
5 |
+
torch
|
6 |
+
gradio
|
7 |
+
fairseq2 --pre --extra-index-url https://fair.pkg.atmeta.com/fairseq2/pt2.1.0/cu121
|