Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -5,7 +5,9 @@ import moviepy.editor as mp
|
|
5 |
from TTS.api import TTS
|
6 |
import torch
|
7 |
import assemblyai as aai
|
|
|
8 |
os.environ["COQUI_TOS_AGREED"] = "1"
|
|
|
9 |
# Download necessary models if not already present
|
10 |
model_files = {
|
11 |
"wav2lip.pth": "https://github.com/justinjohn0306/Wav2Lip/releases/download/models/wav2lip.pth",
|
@@ -15,14 +17,12 @@ model_files = {
|
|
15 |
"s3fd.pth": "https://www.adrianbulat.com/downloads/python-fan/s3fd-619a316812.pth"
|
16 |
}
|
17 |
|
18 |
-
|
19 |
-
|
20 |
device = "cpu"
|
21 |
|
|
|
22 |
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
|
23 |
|
24 |
-
|
25 |
-
|
26 |
for filename, url in model_files.items():
|
27 |
file_path = os.path.join("checkpoints" if "pth" in filename else "face_detection", filename)
|
28 |
if not os.path.exists(file_path):
|
@@ -31,8 +31,6 @@ for filename, url in model_files.items():
|
|
31 |
with open(file_path, 'wb') as f:
|
32 |
f.write(r.content)
|
33 |
|
34 |
-
|
35 |
-
|
36 |
# Translation class
|
37 |
class translation:
|
38 |
def __init__(self, video_path, original_language, target_language):
|
@@ -87,11 +85,11 @@ class translation:
|
|
87 |
translated_text = self.translate_text(transcript_text)
|
88 |
translated_audio_path = self.generate_audio(translated_text)
|
89 |
|
90 |
-
# Run Wav2Lip inference
|
91 |
-
|
|
|
92 |
return 'output_video.mp4'
|
93 |
|
94 |
-
|
95 |
# Gradio Interface
|
96 |
def app(video_path, original_language, target_language):
|
97 |
translator = translation(video_path, original_language, target_language)
|
@@ -108,4 +106,119 @@ interface = gr.Interface(
|
|
108 |
outputs=gr.Video(label="Translated Video")
|
109 |
)
|
110 |
|
111 |
-
interface.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
from TTS.api import TTS
|
6 |
import torch
|
7 |
import assemblyai as aai
|
8 |
+
|
9 |
os.environ["COQUI_TOS_AGREED"] = "1"
|
10 |
+
|
11 |
# Download necessary models if not already present
|
12 |
model_files = {
|
13 |
"wav2lip.pth": "https://github.com/justinjohn0306/Wav2Lip/releases/download/models/wav2lip.pth",
|
|
|
17 |
"s3fd.pth": "https://www.adrianbulat.com/downloads/python-fan/s3fd-619a316812.pth"
|
18 |
}
|
19 |
|
|
|
|
|
20 |
device = "cpu"
|
21 |
|
22 |
+
# Initialize TTS model
|
23 |
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
|
24 |
|
25 |
+
# Download models
|
|
|
26 |
for filename, url in model_files.items():
|
27 |
file_path = os.path.join("checkpoints" if "pth" in filename else "face_detection", filename)
|
28 |
if not os.path.exists(file_path):
|
|
|
31 |
with open(file_path, 'wb') as f:
|
32 |
f.write(r.content)
|
33 |
|
|
|
|
|
34 |
# Translation class
|
35 |
class translation:
|
36 |
def __init__(self, video_path, original_language, target_language):
|
|
|
85 |
translated_text = self.translate_text(transcript_text)
|
86 |
translated_audio_path = self.generate_audio(translated_text)
|
87 |
|
88 |
+
# Run Wav2Lip inference (update the path to inference.py)
|
89 |
+
inference_script_path = "inference.py" # Update this to the actual location of inference.py
|
90 |
+
os.system(f"python {inference_script_path} --checkpoint_path 'checkpoints/wav2lip_gan.pth' --face {self.video_path} --audio {translated_audio_path} --outfile 'output_video.mp4'")
|
91 |
return 'output_video.mp4'
|
92 |
|
|
|
93 |
# Gradio Interface
|
94 |
def app(video_path, original_language, target_language):
|
95 |
translator = translation(video_path, original_language, target_language)
|
|
|
106 |
outputs=gr.Video(label="Translated Video")
|
107 |
)
|
108 |
|
109 |
+
interface.launch()
|
110 |
+
|
111 |
+
|
112 |
+
|
113 |
+
|
114 |
+
# import os
|
115 |
+
# import requests
|
116 |
+
# import gradio as gr
|
117 |
+
# import moviepy.editor as mp
|
118 |
+
# from TTS.api import TTS
|
119 |
+
# import torch
|
120 |
+
# import assemblyai as aai
|
121 |
+
# os.environ["COQUI_TOS_AGREED"] = "1"
|
122 |
+
# # Download necessary models if not already present
|
123 |
+
# model_files = {
|
124 |
+
# "wav2lip.pth": "https://github.com/justinjohn0306/Wav2Lip/releases/download/models/wav2lip.pth",
|
125 |
+
# "wav2lip_gan.pth": "https://github.com/justinjohn0306/Wav2Lip/releases/download/models/wav2lip_gan.pth",
|
126 |
+
# "resnet50.pth": "https://github.com/justinjohn0306/Wav2Lip/releases/download/models/resnet50.pth",
|
127 |
+
# "mobilenet.pth": "https://github.com/justinjohn0306/Wav2Lip/releases/download/models/mobilenet.pth",
|
128 |
+
# "s3fd.pth": "https://www.adrianbulat.com/downloads/python-fan/s3fd-619a316812.pth"
|
129 |
+
# }
|
130 |
+
|
131 |
+
|
132 |
+
|
133 |
+
# device = "cpu"
|
134 |
+
|
135 |
+
# tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
|
136 |
+
|
137 |
+
|
138 |
+
|
139 |
+
# for filename, url in model_files.items():
|
140 |
+
# file_path = os.path.join("checkpoints" if "pth" in filename else "face_detection", filename)
|
141 |
+
# if not os.path.exists(file_path):
|
142 |
+
# print(f"Downloading {filename}...")
|
143 |
+
# r = requests.get(url)
|
144 |
+
# with open(file_path, 'wb') as f:
|
145 |
+
# f.write(r.content)
|
146 |
+
|
147 |
+
|
148 |
+
|
149 |
+
# # Translation class
|
150 |
+
# class translation:
|
151 |
+
# def __init__(self, video_path, original_language, target_language):
|
152 |
+
# self.video_path = video_path
|
153 |
+
# self.original_language = original_language
|
154 |
+
# self.target_language = target_language
|
155 |
+
|
156 |
+
# def org_language_parameters(self, original_language):
|
157 |
+
# language_codes = {'English': 'en', 'German': 'de', 'Italian': 'it', 'Spanish': 'es'}
|
158 |
+
# self.lan_code = language_codes.get(original_language, '')
|
159 |
+
|
160 |
+
# def target_language_parameters(self, target_language):
|
161 |
+
# language_codes = {'English': 'en', 'German': 'de', 'Italian': 'it', 'Spanish': 'es'}
|
162 |
+
# self.tran_code = language_codes.get(target_language, '')
|
163 |
+
|
164 |
+
# def extract_audio(self):
|
165 |
+
# video = mp.VideoFileClip(self.video_path)
|
166 |
+
# audio = video.audio
|
167 |
+
# audio_path = "output_audio.wav"
|
168 |
+
# audio.write_audiofile(audio_path)
|
169 |
+
# return audio_path
|
170 |
+
|
171 |
+
# def transcribe_audio(self, audio_path):
|
172 |
+
# aai.settings.api_key = os.getenv("ASSEMBLYAI_API_KEY")
|
173 |
+
# config = aai.TranscriptionConfig(language_code=self.lan_code)
|
174 |
+
# transcriber = aai.Transcriber(config=config)
|
175 |
+
# transcript = transcriber.transcribe(audio_path)
|
176 |
+
# return transcript.text
|
177 |
+
|
178 |
+
# def translate_text(self, transcript_text):
|
179 |
+
# base_url = "https://api.cognitive.microsofttranslator.com/translate"
|
180 |
+
# headers = {
|
181 |
+
# "Ocp-Apim-Subscription-Key": os.getenv("MICROSOFT_TRANSLATOR_API_KEY"),
|
182 |
+
# "Content-Type": "application/json",
|
183 |
+
# "Ocp-Apim-Subscription-Region": "southeastasia"
|
184 |
+
# }
|
185 |
+
# params = {"api-version": "3.0", "from": self.lan_code, "to": self.tran_code}
|
186 |
+
# body = [{"text": transcript_text}]
|
187 |
+
# response = requests.post(base_url, headers=headers, params=params, json=body)
|
188 |
+
# translation = response.json()[0]["translations"][0]["text"]
|
189 |
+
# return translation
|
190 |
+
|
191 |
+
# def generate_audio(self, translated_text):
|
192 |
+
# tts.tts_to_file(text=translated_text, speaker_wav='output_audio.wav', file_path="output_synth.wav", language=self.tran_code)
|
193 |
+
# return "output_synth.wav"
|
194 |
+
|
195 |
+
# def translate_video(self):
|
196 |
+
# audio_path = self.extract_audio()
|
197 |
+
# self.org_language_parameters(self.original_language)
|
198 |
+
# self.target_language_parameters(self.target_language)
|
199 |
+
# transcript_text = self.transcribe_audio(audio_path)
|
200 |
+
# translated_text = self.translate_text(transcript_text)
|
201 |
+
# translated_audio_path = self.generate_audio(translated_text)
|
202 |
+
|
203 |
+
# # Run Wav2Lip inference
|
204 |
+
# os.system(f"python inference.py --checkpoint_path 'checkpoints/wav2lip_gan.pth' --face {self.video_path} --audio {translated_audio_path} --outfile 'output_video.mp4'")
|
205 |
+
# return 'output_video.mp4'
|
206 |
+
|
207 |
+
|
208 |
+
# # Gradio Interface
|
209 |
+
# def app(video_path, original_language, target_language):
|
210 |
+
# translator = translation(video_path, original_language, target_language)
|
211 |
+
# video_file = translator.translate_video()
|
212 |
+
# return video_file
|
213 |
+
|
214 |
+
# interface = gr.Interface(
|
215 |
+
# fn=app,
|
216 |
+
# inputs=[
|
217 |
+
# gr.Video(label="Video Path"),
|
218 |
+
# gr.Dropdown(["English", "German", "Italian", "Spanish"], label="Original Language"),
|
219 |
+
# gr.Dropdown(["English", "German", "Italian", "Spanish"], label="Targeted Language"),
|
220 |
+
# ],
|
221 |
+
# outputs=gr.Video(label="Translated Video")
|
222 |
+
# )
|
223 |
+
|
224 |
+
# interface.launch()
|