uasername commited on
Commit
20d75d2
·
verified ·
1 Parent(s): c2a302a

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +31 -19
app.py CHANGED
@@ -14,43 +14,38 @@ import soundfile as sf
14
  import os
15
  import numpy as np
16
 
 
 
17
 
18
  # Initialize the Kokoro pipeline
19
  pipeline = KPipeline(lang_code='a') # 'a' stands for American English
20
 
 
 
 
21
  @tool
22
  def text_to_speech_kokoro(text: str, voice: str = 'af_heart', speed: float = 1.0) -> str:
23
- """Convert text to speech using the Kokoro-82M model.
24
-
25
- Args:
26
- text: The text to be converted to speech.
27
- voice: The voice to use for speech synthesis (default is 'af_heart').
28
- speed: The speed of the speech (default is 1.0).
29
-
30
- Returns:
31
- An AgentAudio object with the relative URL to the generated audio file.
32
- """
33
  try:
34
- # Generate speech audio
35
  generator = pipeline(text, voice=voice, speed=speed, split_pattern=r'\n+')
36
  audio_segments = []
37
  for _, _, audio in generator:
38
  audio_segments.append(audio)
 
39
  if not audio_segments:
40
  raise ValueError("No audio generated.")
41
- # Concatenate segments into one audio array
42
  full_audio = np.concatenate(audio_segments)
43
  sample_rate = 24000 # Kokoro outputs at 24 kHz
44
- # Ensure the static folder exists and save the file there
45
  os.makedirs("tools", exist_ok=True)
46
- filename = os.path.join("tools", "output.wav")
47
- sf.write(filename, full_audio, sample_rate)
48
- # Return an AgentAudio object pointing to the relative URL of the audio file
49
- from smolagents.agent_types import AgentAudio
50
- return AgentAudio(f"tools/output.wav")
51
  except Exception as e:
52
  return f"Error generating speech: {str(e)}"
53
 
 
54
  @tool
55
  def search_dad_jokes(term: str) -> str:
56
  """A tool that searches for dad jokes containing a specific term.
@@ -111,4 +106,21 @@ agent = CodeAgent(
111
  )
112
 
113
 
114
- GradioUI(agent).launch()
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  import os
15
  import numpy as np
16
 
17
+ import gradio as gr
18
+
19
 
20
  # Initialize the Kokoro pipeline
21
  pipeline = KPipeline(lang_code='a') # 'a' stands for American English
22
 
23
+ # Define audio output path
24
+ AUDIO_OUTPUT_PATH = "tools/output.wav"
25
+
26
  @tool
27
  def text_to_speech_kokoro(text: str, voice: str = 'af_heart', speed: float = 1.0) -> str:
28
+ """Convert text to speech using Kokoro-82M model and return audio file path."""
 
 
 
 
 
 
 
 
 
29
  try:
 
30
  generator = pipeline(text, voice=voice, speed=speed, split_pattern=r'\n+')
31
  audio_segments = []
32
  for _, _, audio in generator:
33
  audio_segments.append(audio)
34
+
35
  if not audio_segments:
36
  raise ValueError("No audio generated.")
37
+
38
  full_audio = np.concatenate(audio_segments)
39
  sample_rate = 24000 # Kokoro outputs at 24 kHz
40
+
41
  os.makedirs("tools", exist_ok=True)
42
+ sf.write(AUDIO_OUTPUT_PATH, full_audio, sample_rate)
43
+
44
+ return AUDIO_OUTPUT_PATH
 
 
45
  except Exception as e:
46
  return f"Error generating speech: {str(e)}"
47
 
48
+
49
  @tool
50
  def search_dad_jokes(term: str) -> str:
51
  """A tool that searches for dad jokes containing a specific term.
 
106
  )
107
 
108
 
109
+ #GradioUI(agent).launch()
110
+
111
+ # Gradio wrapper function
112
+ def gradio_text_to_speech(text):
113
+ audio_path = text_to_speech_kokoro(text)
114
+ return audio_path if os.path.exists(audio_path) else "Error: No audio file generated."
115
+
116
+ # Define the Gradio UI
117
+ with gr.Blocks() as demo:
118
+ gr.Markdown("### Text-to-Speech with Kokoro AI 🎙️")
119
+ with gr.Row():
120
+ input_box = gr.Textbox(label="Enter text")
121
+ output_audio = gr.Audio(label="Generated Speech", type="filepath")
122
+
123
+ btn = gr.Button("Generate Speech")
124
+ btn.click(gradio_text_to_speech, inputs=input_box, outputs=output_audio)
125
+
126
+ demo.launch()