File size: 8,710 Bytes
db31449 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 |
import gradio as gr
import os
import torch
from openvoice import se_extractor
from openvoice.api import ToneColorConverter
import whisper
from moviepy.editor import VideoFileClip
from pydub import AudioSegment
from df.enhance import enhance, init_df, load_audio, save_audio
import translators as ts
from melo.api import TTS
from concurrent.futures import ThreadPoolExecutor
import ffmpeg
def process_video(video_file, language_choice):
if video_file == None or language_choice == None:
return None
# Initialize paths and devices
ckpt_converter = 'checkpoints_v2/converter'
device = "cuda:0" if torch.cuda.is_available() else "cpu"
output_dir = 'outputs_v2'
os.makedirs(output_dir, exist_ok=True)
tone_color_converter = ToneColorConverter(f'{ckpt_converter}/config.json', device=device)
tone_color_converter.load_ckpt(f'{ckpt_converter}/checkpoint.pth')
# Process the reference video
reference_video = VideoFileClip(video_file)
reference_audio = os.path.join(output_dir, "reference_audio.wav")
reference_video.audio.write_audiofile(reference_audio)
audio = AudioSegment.from_file(reference_audio)
resampled_audio = audio.set_frame_rate(48000)
resampled_audio.export(reference_audio, format="wav")
# Enhance the audio
model, df_state, _ = init_df()
audio, _ = load_audio(reference_audio, sr=df_state.sr())
enhanced = enhance(model, df_state, audio)
save_audio(reference_audio, enhanced, df_state.sr())
reference_speaker = reference_audio # This is the voice you want to clone
target_se, audio_name = se_extractor.get_se(reference_speaker, tone_color_converter, vad=False)
src_path = os.path.join(output_dir, "tmp.wav")
# Speed is adjustable
speed = 1.0
# Transcribe the original audio with timestamps
sttmodel = whisper.load_model("base")
sttresult = sttmodel.transcribe(reference_speaker, verbose=True)
# Print the original transcription
print(sttresult["text"])
print(sttresult["language"])
# Get the segments with start and end times
segments = sttresult['segments']
# Choose the target language for translation
language = 'EN_NEWEST'
match language_choice:
case 'en':
language = 'EN_NEWEST'
case 'es':
language = 'ES'
case 'fr':
language = 'FR'
case 'zh':
language = 'ZH'
case 'ja':
language = 'JP'
case 'ko':
language = 'KR'
case _:
language = 'EN_NEWEST'
# Translate the transcription segment by segment
def translate_segment(segment):
return segment["start"], segment["end"], ts.translate_text(query_text=segment["text"], translator="google", to_language=language_choice)
# Batch translation to reduce memory load
batch_size = 2
translation_segments = []
for i in range(0, len(segments), batch_size):
batch = segments[i:i + batch_size]
with ThreadPoolExecutor(max_workers=5) as executor:
batch_translations = list(executor.map(translate_segment, batch))
translation_segments.extend(batch_translations)
# Generate the translated audio for each segment
model = TTS(language=language, device=device)
speaker_ids = model.hps.data.spk2id
def generate_segment_audio(segment, speaker_id):
start, end, translated_text = segment
segment_path = os.path.join(output_dir, f'segment_{start}_{end}.wav')
model.tts_to_file(translated_text, speaker_id, segment_path, speed=speed)
return segment_path, start, end, translated_text
for speaker_key in speaker_ids.keys():
speaker_id = speaker_ids[speaker_key]
speaker_key = speaker_key.lower().replace('_', '-')
source_se = torch.load(f'checkpoints_v2/base_speakers/ses/{speaker_key}.pth', map_location=device)
segment_files = []
subtitle_entries = []
for segment in translation_segments:
segment_file, start, end, translated_text = generate_segment_audio(segment, speaker_id)
# Run the tone color converter
encode_message = "@MyShell"
tone_color_converter.convert(
audio_src_path=segment_file,
src_se=source_se,
tgt_se=target_se,
output_path=segment_file,
message=encode_message)
segment_files.append((segment_file, start, end, translated_text))
# Combine the audio segments
combined_audio = AudioSegment.empty()
video_segments = []
previous_end = 0
subtitle_counter = 1
for segment_file, start, end, translated_text in segment_files:
segment_audio = AudioSegment.from_file(segment_file)
combined_audio += segment_audio
# Calculate the duration of the audio segment
audio_duration = len(segment_audio) / 1000.0
# Add the subtitle entry for this segment
subtitle_entries.append((subtitle_counter, previous_end, previous_end + audio_duration, translated_text))
subtitle_counter += 1
# Get the corresponding video segment and adjust its speed to match the audio duration
video_segment = (
ffmpeg
.input(reference_video.filename, ss=start, to=end)
.filter('setpts', f'PTS / {(end - start) / audio_duration}')
)
video_segments.append((video_segment, ffmpeg.input(segment_file)))
previous_end += audio_duration
save_path = os.path.join(output_dir, f'output_v2_{speaker_key}.wav')
combined_audio.export(save_path, format="wav")
# Combine video and audio segments using ffmpeg
video_and_audio_files = [item for sublist in video_segments for item in sublist]
joined = (
ffmpeg
.concat(*video_and_audio_files, v=1, a=1)
.node
)
final_video_path = os.path.join(output_dir, f'final_video_{speaker_key}.mp4')
try:
(
ffmpeg
.output(joined[0], joined[1], final_video_path, vcodec='libx264', acodec='aac')
.run(overwrite_output=True)
)
except ffmpeg.Error as e:
print('ffmpeg error:', e)
print(e.stderr.decode('utf-8'))
print(f"Final video without subtitles saved to: {final_video_path}")
# Generate subtitles file in SRT format
srt_path = os.path.join(output_dir, 'subtitles.srt')
with open(srt_path, 'w', encoding='utf-8') as srt_file:
for entry in subtitle_entries:
index, start, end, text = entry
start_hours, start_minutes = divmod(int(start), 3600)
start_minutes, start_seconds = divmod(start_minutes, 60)
start_milliseconds = int((start * 1000) % 1000)
end_hours, end_minutes = divmod(int(end), 3600)
end_minutes, end_seconds = divmod(end_minutes, 60)
end_milliseconds = int((end * 1000) % 1000)
srt_file.write(f"{index}\n")
srt_file.write(f"{start_hours:02}:{start_minutes:02}:{start_seconds:02},{start_milliseconds:03} --> "
f"{end_hours:02}:{end_minutes:02}:{end_seconds:02},{end_milliseconds:03}\n")
srt_file.write(f"{text}\n\n")
# Add subtitles to the video
final_video_with_subs_path = os.path.join(output_dir, f'final_video_with_subs_{speaker_key}.mp4')
try:
(
ffmpeg
.input(final_video_path)
.output(final_video_with_subs_path, vf=f"subtitles={srt_path}")
.run(overwrite_output=True)
)
except ffmpeg.Error as e:
print('ffmpeg error:', e)
print(e.stderr.decode('utf-8'))
print(f"Final video with subtitles saved to: {final_video_with_subs_path}")
return final_video_with_subs_path
# Define Gradio interface
def gradio_interface(video_file, language_choice):
return process_video(video_file, language_choice)
language_choices = ts.get_languages("google")["en"]
gr.Interface(
fn=gradio_interface,
inputs=[
gr.Video(label="Upload Video"),
gr.Dropdown(choices=language_choices, label="Choose Language for Translation")
],
outputs=gr.Video(label="Translated Video"),
title="Video Translation and Voice Cloning",
description="Upload a video, choose a language to translate the audio, and download the processed video with translated audio."
).launch()
|