j commited on
Commit
cc9ce25
1 Parent(s): e9be68c

convert video inputs and remove HARP stuff

Browse files
Files changed (1) hide show
  1. app.py +20 -12
app.py CHANGED
@@ -4,10 +4,17 @@ from audioldm import build_model, text_to_audio
4
  import gradio as gr
5
  import soundfile as sf
6
  from datetime import datetime
 
 
 
7
 
8
  audioldm = build_model(model_name="audioldm-l-full")
9
 
10
  def process_fn(input_audio_path, seed, guidance_scale, num_inference_steps, num_candidates, audio_length_in_s):
 
 
 
 
11
  waveform = text_to_audio(
12
  audioldm,
13
  'placeholder',
@@ -26,14 +33,17 @@ def process_fn(input_audio_path, seed, guidance_scale, num_inference_steps, num_
26
  return filename
27
 
28
 
29
- card = ModelCard(
30
- name='AudioLDM Variations',
31
- description='AudioLDM Variation Generator, operates on region selected in track.',
32
- author='Team Audio',
33
- tags=['AudioLDM', 'Variations', 'audio-to-audio']
34
- )
 
 
35
 
36
- with gr.Blocks() as webapp:
 
37
  # Define your Gradio interface
38
  inputs = [
39
  gr.Audio(
@@ -66,13 +76,11 @@ with gr.Blocks() as webapp:
66
  minimum=2.5, maximum=10.0,
67
  step=2.5, value=5,
68
  label="Duration"
69
- ),
70
- ]
71
 
72
  output = gr.Audio(label="Audio Output", type="filepath", format="wav", elem_id="audio")
 
73
 
74
- ctrls_data, ctrls_button, process_button, cancel_button = build_endpoint(inputs, output, process_fn, card)
75
-
76
- # queue the webapp: https://www.gradio.app/guides/setting-up-a-demo-for-maximum-performance
77
  webapp.queue()
78
  webapp.launch()
 
4
  import gradio as gr
5
  import soundfile as sf
6
  from datetime import datetime
7
+ import subprocess
8
+ import os
9
+ import sys
10
 
11
  audioldm = build_model(model_name="audioldm-l-full")
12
 
13
  def process_fn(input_audio_path, seed, guidance_scale, num_inference_steps, num_candidates, audio_length_in_s):
14
+ video_extensions = (".mp4", ".avi", ".mkv", ".flv", ".mov", ".wmv", ".webm")
15
+ if input_audio_path.lower().endswith(video_extensions):
16
+ input_audio_path = convert_video_to_audio_ffmpeg(input_audio_path)
17
+
18
  waveform = text_to_audio(
19
  audioldm,
20
  'placeholder',
 
33
  return filename
34
 
35
 
36
+ def convert_video_to_audio_ffmpeg(video_file, output_ext="wav"):
37
+ """Converts video to audio directly using `ffmpeg` command
38
+ with the help of subprocess module"""
39
+ filename, ext = os.path.splitext(video_file)
40
+ subprocess.call(["ffmpeg", "-y", "-i", video_file, f"{filename}.{output_ext}"],
41
+ stdout=subprocess.DEVNULL,
42
+ stderr=subprocess.STDOUT)
43
+ return f"{filename}.{output_ext}"
44
 
45
+ webapp = gr.Interface(
46
+ fn = process_fn,
47
  # Define your Gradio interface
48
  inputs = [
49
  gr.Audio(
 
76
  minimum=2.5, maximum=10.0,
77
  step=2.5, value=5,
78
  label="Duration"
79
+ )
80
+ ],
81
 
82
  output = gr.Audio(label="Audio Output", type="filepath", format="wav", elem_id="audio")
83
+ )
84
 
 
 
 
85
  webapp.queue()
86
  webapp.launch()