Skip to content

FunASR-powered video transcription, subtitle generation, and LLM-assisted clipping tool with a local Gradio UI.

support output_dir

维石210838065ff9
diff --git a/funclip/launch.py b/funclip/launch.py
index f3915ce..bae7e4a 100644
--- a/funclip/launch.py
+++ b/funclip/launch.py
@@ -1,3 +1,4 @@
+import os
import gradio as gr
from funasr import AutoModel
from videoclipper import VideoClipper
@@ -12,34 +13,58 @@ if __name__ == "__main__":
)
audio_clipper = VideoClipper(funasr_model)
- def audio_recog(audio_input, sd_switch, hotwords):
- return audio_clipper.recog(audio_input, sd_switch, hotwords=hotwords)
+ def audio_recog(audio_input, sd_switch, hotwords, output_dir):
+ return audio_clipper.recog(audio_input, sd_switch, hotwords, output_dir=output_dir)
- def video_recog(video_input, sd_switch, hotwords):
- return audio_clipper.video_recog(video_input, sd_switch, hotwords)
+ def video_recog(video_input, sd_switch, hotwords, output_dir):
+ return audio_clipper.video_recog(video_input, sd_switch, hotwords, output_dir=output_dir)
- def video_clip(dest_text, video_spk_input, start_ost, end_ost, state):
- return audio_clipper.video_clip(dest_text, start_ost, end_ost, state, dest_spk=video_spk_input)
+ def video_clip(dest_text, video_spk_input, start_ost, end_ost, state, output_dir):
+ return audio_clipper.video_clip(
+ dest_text, start_ost, end_ost, state, dest_spk=video_spk_input, output_dir=output_dir
+ )
- def video_clip_addsub(dest_text, video_spk_input, start_ost, end_ost, state, font_size, font_color):
- return audio_clipper.video_clip(dest_text, start_ost, end_ost, state, font_size, font_color, add_sub=True, dest_spk=video_spk_input)
-
- def mix_recog(video_input, audio_input, sd_switch, hotwords):
+ def mix_recog(video_input, audio_input, sd_switch, hotwords, output_dir):
+ # output_dir here is for funasr model generate
+ # both relative and absolute paths are OK
+ output_dir = output_dir.strip()
+ if not len(output_dir):
+ output_dir = None
+ else:
+ output_dir = os.path.abspath(output_dir)
audio_state, video_state = None, None
if video_input is not None:
- res_text, res_srt, video_state = video_recog(video_input, sd_switch, hotwords)
+ res_text, res_srt, video_state = video_recog(
+ video_input, sd_switch, hotwords, output_dir=output_dir)
return res_text, res_srt, video_state, None
if audio_input is not None:
- res_text, res_srt, audio_state = audio_recog(audio_input, sd_switch, hotwords)
+ res_text, res_srt, audio_state = audio_recog(
+ audio_input, sd_switch, hotwords, output_dir=output_dir)
return res_text, res_srt, None, audio_state
- def mix_clip(dest_text, video_spk_input, start_ost, end_ost, video_state, audio_state):
+ def mix_clip(dest_text, video_spk_input, start_ost, end_ost, video_state, audio_state, output_dir):
+ # output_dir here is for moviepy, vedio saving
+ # supposed to be absolute path only
+ output_dir = output_dir.strip()
+ if not len(output_dir):
+ output_dir = None
+ else:
+ output_dir = os.path.abspath(output_dir)
if video_state is not None:
- clip_video_file, message, clip_srt = audio_clipper.video_clip(dest_text, start_ost, end_ost, video_state, dest_spk=video_spk_input)
+ clip_video_file, message, clip_srt = audio_clipper.video_clip(
+ dest_text, start_ost, end_ost, video_state, dest_spk=video_spk_input, output_dir=output_dir)
return clip_video_file, None, message, clip_srt
if audio_state is not None:
- (sr, res_audio), message, clip_srt = audio_clipper.clip(dest_text, start_ost, end_ost, audio_state, dest_spk=video_spk_input)
+ (sr, res_audio), message, clip_srt = audio_clipper.clip(
+ dest_text, start_ost, end_ost, audio_state, dest_spk=video_spk_input, output_dir=output_dir)
return None, (sr, res_audio), message, clip_srt
+
+ def video_clip_addsub(dest_text, video_spk_input, start_ost, end_ost, state, output_dir, font_size, font_color):
+ return audio_clipper.video_clip(
+ dest_text, start_ost, end_ost, state,
+ font_size=font_size, font_color=font_color,
+ add_sub=True, dest_spk=video_spk_input, output_dir=output_dir
+ )
# gradio interface
with gr.Blocks() as funclip_service:
@@ -68,6 +93,7 @@ if __name__ == "__main__":
with gr.Row():
video_sd_switch = gr.Radio(["No", "Yes"], label="👥是否区分说话人 Recognize Speakers", value='No')
hotwords_input = gr.Textbox(label="🚒热词 Hotwords")
+ output_dir = gr.Textbox(label="文件输出路径 File Output Dir", value=" ")
recog_button = gr.Button("👂识别 Recognize")
video_text_output = gr.Textbox(label="✏️识别结果 Recognition Result")
video_srt_output = gr.Textbox(label="📖SRT字幕内容 RST Subtitles")
@@ -90,13 +116,33 @@ if __name__ == "__main__":
srt_clipped = gr.Textbox(label="📖裁剪部分SRT字幕内容 Clipped RST Subtitles")
recog_button.click(mix_recog,
- inputs=[video_input, audio_input, video_sd_switch, hotwords_input],
+ inputs=[video_input,
+ audio_input,
+ video_sd_switch,
+ hotwords_input,
+ output_dir,
+ ],
outputs=[video_text_output, video_srt_output, video_state, audio_state])
clip_button.click(mix_clip,
- inputs=[video_text_input, video_spk_input, video_start_ost, video_end_ost, video_state, audio_state],
+ inputs=[video_text_input,
+ video_spk_input,
+ video_start_ost,
+ video_end_ost,
+ video_state,
+ audio_state,
+ output_dir
+ ],
outputs=[video_output, audio_output, clip_message, srt_clipped])
clip_subti_button.click(video_clip_addsub,
- inputs=[video_text_input, video_spk_input, video_start_ost, video_end_ost, video_state, font_size, font_color],
+ inputs=[video_text_input,
+ video_spk_input,
+ video_start_ost,
+ video_end_ost,
+ video_state,
+ output_dir,
+ font_size,
+ font_color,
+ ],
outputs=[video_output, clip_message, srt_clipped])
# start gradio service in local
diff --git a/funclip/videoclipper.py b/funclip/videoclipper.py
index b772685..cdbcae3 100644
--- a/funclip/videoclipper.py
+++ b/funclip/videoclipper.py
@@ -21,7 +21,7 @@ class VideoClipper():
self.funasr_model = funasr_model
self.GLOBAL_COUNT = 0
- def recog(self, audio_input, sd_switch='no', state=None, hotwords=""):
+ def recog(self, audio_input, sd_switch='no', state=None, hotwords="", output_dir=None):
if state is None:
state = {}
sr, data = audio_input
@@ -46,7 +46,8 @@ class VideoClipper():
sentence_timestamp=True,
return_raw_text=True,
is_final=True,
- hotword=hotwords)
+ hotword=hotwords,
+ output_dir=output_dir,)
res_srt = generate_srt(rec_result[0]['sentence_info'])
state['recog_res_raw'] = rec_result[0]['raw_text']
state['timestamp'] = rec_result[0]['timestamp']
@@ -54,7 +55,7 @@ class VideoClipper():
res_text = rec_result[0]['text']
return res_text, res_srt, state
- def clip(self, dest_text, start_ost, end_ost, state, dest_spk=None):
+ def clip(self, dest_text, start_ost, end_ost, state, dest_spk=None, output_dir=None):
# get from state
audio_input = state['audio_input']
recog_res_raw = state['recog_res_raw']
@@ -116,12 +117,19 @@ class VideoClipper():
res_audio = data
return (sr, res_audio), message, clip_srt
- def video_recog(self, video_filename, sd_switch='no', hotwords=""):
+ def video_recog(self, video_filename, sd_switch='no', hotwords="", output_dir=None):
video = mpy.VideoFileClip(video_filename)
# Extract the base name, add '_clip.mp4', and 'wav'
- base_name, _ = os.path.splitext(video_filename)
- clip_video_file = base_name + '_clip.mp4'
- audio_file = base_name + '.wav'
+ if output_dir is not None:
+ _, base_name = os.path.split(video_filename)
+ base_name, _ = os.path.splitext(base_name)
+ clip_video_file = base_name + '_clip.mp4'
+ audio_file = base_name + '.wav'
+ audio_file = os.path.join(output_dir, audio_file)
+ else:
+ base_name, _ = os.path.splitext(video_filename)
+ clip_video_file = base_name + '_clip.mp4'
+ audio_file = base_name + '.wav'
video.audio.write_audiofile(audio_file)
wav = librosa.load(audio_file, sr=16000)[0]
# delete the audio file after processing
@@ -133,9 +141,18 @@ class VideoClipper():
'video': video,
}
# res_text, res_srt = self.recog((16000, wav), state)
- return self.recog((16000, wav), sd_switch, state, hotwords)
+ return self.recog((16000, wav), sd_switch, state, hotwords, output_dir)
- def video_clip(self, dest_text, start_ost, end_ost, state, font_size=32, font_color='white', add_sub=False, dest_spk=None):
+ def video_clip(self,
+ dest_text,
+ start_ost,
+ end_ost,
+ state,
+ font_size=32,
+ font_color='white',
+ add_sub=False,
+ dest_spk=None,
+ output_dir=None):
# get from state
recog_res_raw = state['recog_res_raw']
timestamp = state['timestamp']
@@ -209,8 +226,18 @@ class VideoClipper():
logging.warning("Concating...")
if len(concate_clip) > 1:
video_clip = concatenate_videoclips(concate_clip)
- clip_video_file = clip_video_file[:-4] + '_no{}.mp4'.format(self.GLOBAL_COUNT)
- video_clip.write_videofile(clip_video_file, audio_codec="aac", temp_audiofile="video_no{}.mp4".format(self.GLOBAL_COUNT))
+ # clip_video_file = clip_video_file[:-4] + '_no{}.mp4'.format(self.GLOBAL_COUNT)
+ if output_dir is not None:
+ os.makedirs(output_dir, exist_ok=True)
+ _, file_with_extension = os.path.split(clip_video_file)
+ clip_video_file_name, _ = os.path.splitext(file_with_extension)
+ print(output_dir, clip_video_file)
+ clip_video_file = os.path.join(output_dir, "{}_no{}.mp4".format(clip_video_file_name, self.GLOBAL_COUNT))
+ temp_audio_file = os.path.join(output_dir, "{}_no{}.mp3".format(clip_video_file_name, self.GLOBAL_COUNT))
+ else:
+ clip_video_file = clip_video_file[:-4] + '_no{}.mp4'.format(self.GLOBAL_COUNT)
+ temp_audio_file = clip_video_file[:-4] + '_tempaudio_no{}.mp4'.format(self.GLOBAL_COUNT)
+ video_clip.write_videofile(clip_video_file, audio_codec="aac", temp_audiofile=temp_audio_file)
self.GLOBAL_COUNT += 1
else:
clip_video_file = video_filename