SignLanguage / audio.py
ZiyuG's picture
Update audio.py
ad8b014 verified
raw
history blame
4.65 kB
from moviepy.editor import *
from gtts import gTTS
from pydub import AudioSegment
import tempfile
import os
import pyttsx3
from ms_tts import ms_tts_gen
os.environ["IMAGEIO_FFMPEG_EXE"] = "/usr/local/bin/ffmpeg"
# os.environ["IMAGEMAGICK_BINARY"] = "/usr/local/bin/convert"
def gen_audio(subtitles, tmpdir):
# subtitles = [
# [[0, 3], "這裡請注意右臂"],
# [[4, 8], "這裡請注意左臂"],
# [[9, 12], "這裡請注意左手細節"]
# ])
subtitles = merge_subtitles(subtitles)
print("===> Subtitles:")
for subtitle in subtitles:
print(subtitle)
combined_clip = VideoFileClip(tmpdir + "/output.mp4")
combined_clip = combined_clip.resize(width=800)
total_duration = combined_clip.duration
def add_subtitles(clip, subtitles):
txt_clips = []
for start_end, text in subtitles:
start_time, end_time = start_end
duration = end_time - start_time
txt_clip = (TextClip(text, fontsize=clip.w//20, color='white', font='bold.ttf', method='caption', size=(clip.w * 0.9, None))
.set_duration(duration)
.set_position(('center', 4 * clip.h // 5))
.set_start(start_time))
txt_clips.append(txt_clip)
return CompositeVideoClip([clip, *txt_clips])
if len(subtitles) != 0:
video_with_subtitles = combined_clip #add_subtitles(combined_clip, subtitles)
else:
combined_clip.write_videofile(tmpdir + '/output_with_subtitles.mp4', fps=24)
print("Generated Video with Subtitles to:", tmpdir + '/output_with_subtitles.mp4')
return tmpdir + '/output_with_subtitles.mp4'
def generate_audio(subtitles, total_duration):
temp_audio_path = tempfile.mkdtemp()
clips = []
# engine = pyttsx3.init()
# engine.setProperty('voice', 'yue') # yue:粤语 cnm:普通话
# engine.setProperty('rate', 150) # 设置语速
# engine.setProperty('volume', 1.0) # 设置音量
for i, (start_end, text) in enumerate(subtitles):
start_time, end_time = start_end
duration = end_time - start_time
# tts = gTTS(text=text, lang='yue')
audio_path = os.path.join(temp_audio_path, f'subtitle_{i}.wav')
# tts.save(audio_path)
ms_tts_gen(text, audio_path)
audio_segment = AudioSegment.from_wav(audio_path)
new_audio_path = os.path.join(temp_audio_path, f'subtitle_{i}_adjusted.wav')
audio_segment.export(new_audio_path, format="wav")
audio_clip = AudioFileClip(new_audio_path).set_start(start_time).set_duration(min(duration, audio_segment.duration_seconds))
clips.append(audio_clip)
final_audio = CompositeAudioClip(clips)
return final_audio
audio_clip = generate_audio(subtitles, total_duration)
video_with_audio = video_with_subtitles.set_audio(audio_clip)
video_with_audio.write_videofile(tmpdir + '/output_with_subtitles.mp4', fps=24)
print("Generated Video with Subtitles to:", tmpdir + '/output_with_subtitles.mp4')
return tmpdir + '/output_with_subtitles.mp4'
def merge_subtitles(subtitles):
merged_subtitles = []
i = 0
n = len(subtitles)
while i < n:
current_start, current_end = subtitles[i][0]
current_content = subtitles[i][1]
duration = current_end - current_start
i += 1
while duration < 3:
if i >= n:
current_end = current_start + 3
duration = current_end - current_start
break
next_start, next_end = subtitles[i][0]
next_content = subtitles[i][1]
if next_start == current_end:
current_end = next_end
current_content += "、" + next_content
duration = current_end - current_start
i += 1
else:
if duration < 3:
current_end = current_start + 3
duration = current_end - current_start
break
current_content = '這裡請注意' + '、'.join(list(set(current_content.replace("細節", "").split('、'))))
current_content = current_content.replace("左手、右手", "雙手").replace("右手、左手", "雙手")
current_content = current_content.replace("左臂、右臂", "雙臂").replace("右臂、左臂", "雙臂")
merged_subtitles.append([[current_start, current_end], current_content])
return merged_subtitles