lshzhm commited on
Commit
8a3c983
·
1 Parent(s): b61a5a8
F5-TTS/src/f5_tts/infer/infer_cli_test.py CHANGED
@@ -257,6 +257,8 @@ sway_sampling_coef = args.sway_sampling_coef or config.get("sway_sampling_coef",
257
  speed = args.speed or config.get("speed", speed)
258
  fix_duration = args.fix_duration or config.get("fix_duration", fix_duration)
259
 
 
 
260
 
261
  # patches for pip pkg user
262
  if "infer/examples/" in ref_audio:
 
257
  speed = args.speed or config.get("speed", speed)
258
  fix_duration = args.fix_duration or config.get("fix_duration", fix_duration)
259
 
260
+ print("############nfe_step", nfe_step)
261
+
262
 
263
  # patches for pip pkg user
264
  if "infer/examples/" in ref_audio:
app.py CHANGED
@@ -35,7 +35,7 @@ log = logging.getLogger()
35
 
36
 
37
  #@spaces.GPU(duration=120)
38
- def video_to_audio_and_speech(video: gr.Video, prompt: str, text: str, audio_prompt: gr.Audio, text_prompt: str):
39
 
40
  video_path = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4').name
41
 
@@ -65,15 +65,15 @@ def video_to_audio_and_speech(video: gr.Video, prompt: str, text: str, audio_pro
65
  shutil.copy(audio_prompt, audio_p_path)
66
 
67
  if prompt == "":
68
- command = "cd ./MMAudio; python ./demo.py --variant small_44k --output %s --video %s --calc_energy 1" % (output_dir, video_path)
69
  else:
70
- command = "cd ./MMAudio; python ./demo.py --variant small_44k --output %s --video %s --prompt %s --calc_energy 1" % (output_dir, video_path, prompt)
71
  print("v2a command", command)
72
  os.system(command)
73
 
74
  video_gen = video_save_path[:-4]+".mp4.gen.mp4"
75
 
76
- command = "python ./F5-TTS/src/f5_tts/infer/infer_cli_test.py --output_dir %s --start 0 --end 1 --ckpt_file ./F5-TTS/ckpts/v2c/v2c_s44.pt --v2a_path %s --wav_p %s --txt_p \"%s\" --video %s --v2a_wav %s --txt \"%s\"" % (output_dir, output_dir, audio_p_path, text_prompt, video_save_path, video_save_path[:-4]+".flac", text)
77
  print("v2s command", command, video_gen)
78
  os.system(command)
79
 
@@ -89,9 +89,11 @@ video_to_audio_and_speech_tab = gr.Interface(
89
  inputs=[
90
  gr.Video(label="Input Video"),
91
  gr.Text(label='Video-to-Audio Text Prompt'),
 
92
  gr.Text(label='Video-to-Speech Transcription'),
93
  gr.Audio(label='Video-to-Speech Speech Prompt'),
94
  gr.Text(label='Video-to-Speech Speech Prompt Transcription'),
 
95
  ],
96
  outputs=[
97
  gr.Video(label="Video-to-Audio Output"),
@@ -103,16 +105,20 @@ video_to_audio_and_speech_tab = gr.Interface(
103
  [
104
  './tests/0235.mp4',
105
  '',
 
106
  "Who finally decided to show up for work Yay",
107
  './tests/Gobber-00-0778.wav',
108
  "I've still got a few knocking around in here",
 
109
  ],
110
  [
111
  './tests/0778.mp4',
112
  '',
 
113
  "I've still got a few knocking around in here",
114
  './tests/Gobber-00-0235.wav',
115
  "Who finally decided to show up for work Yay",
 
116
  ],
117
  ])
118
 
 
35
 
36
 
37
  #@spaces.GPU(duration=120)
38
+ def video_to_audio_and_speech(video: gr.Video, prompt: str, v2a_num_steps: int, text: str, audio_prompt: gr.Audio, text_prompt: str, v2s_num_steps: int):
39
 
40
  video_path = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4').name
41
 
 
65
  shutil.copy(audio_prompt, audio_p_path)
66
 
67
  if prompt == "":
68
+ command = "cd ./MMAudio; python ./demo.py --variant small_44k --output %s --video %s --calc_energy 1 --num_steps %d" % (output_dir, video_path, v2a_num_steps)
69
  else:
70
+ command = "cd ./MMAudio; python ./demo.py --variant small_44k --output %s --video %s --prompt %s --calc_energy 1 --num_steps %d" % (output_dir, video_path, prompt, v2a_num_steps)
71
  print("v2a command", command)
72
  os.system(command)
73
 
74
  video_gen = video_save_path[:-4]+".mp4.gen.mp4"
75
 
76
+ command = "python ./F5-TTS/src/f5_tts/infer/infer_cli_test.py --output_dir %s --start 0 --end 1 --ckpt_file ./F5-TTS/ckpts/v2c/v2c_s44.pt --v2a_path %s --wav_p %s --txt_p \"%s\" --video %s --v2a_wav %s --txt \"%s\" --nfe_step %d" % (output_dir, output_dir, audio_p_path, text_prompt, video_save_path, video_save_path[:-4]+".flac", text, v2s_num_steps)
77
  print("v2s command", command, video_gen)
78
  os.system(command)
79
 
 
89
  inputs=[
90
  gr.Video(label="Input Video"),
91
  gr.Text(label='Video-to-Audio Text Prompt'),
92
+ gr.Number(label='Video-to-Audio Num Steps', value=25, precision=0, minimum=1),
93
  gr.Text(label='Video-to-Speech Transcription'),
94
  gr.Audio(label='Video-to-Speech Speech Prompt'),
95
  gr.Text(label='Video-to-Speech Speech Prompt Transcription'),
96
+ gr.Number(label='Video-to-Speech Num Steps', value=32, precision=0, minimum=1),
97
  ],
98
  outputs=[
99
  gr.Video(label="Video-to-Audio Output"),
 
105
  [
106
  './tests/0235.mp4',
107
  '',
108
+ 25,
109
  "Who finally decided to show up for work Yay",
110
  './tests/Gobber-00-0778.wav',
111
  "I've still got a few knocking around in here",
112
+ 32,
113
  ],
114
  [
115
  './tests/0778.mp4',
116
  '',
117
+ 25,
118
  "I've still got a few knocking around in here",
119
  './tests/Gobber-00-0235.wav',
120
  "Who finally decided to show up for work Yay",
121
+ 32,
122
  ],
123
  ])
124