File size: 8,843 Bytes
3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 3ac5c08 03eec30 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 |
# filepath: /Users/udaylunawat/Downloads/Data-Science-Projects/NotebookLM_clone/gradio_app.py
import os
import tempfile
import gradio as gr
from notebook_lm_kokoro import generate_podcast_script, KPipeline
import soundfile as sf
import numpy as np
import ast
import shutil
import warnings
import os
import gradio as gr
from notebook_lm_kokoro import generate_podcast_script, generate_audio_from_script
warnings.filterwarnings("ignore")
# A modified version of generate_audio_from_script to accept voice mapping
def generate_audio_from_script_with_voices(script, speaker1_voice, speaker2_voice, output_file):
voice_map = {"Speaker 1": speaker1_voice, "Speaker 2": speaker2_voice}
# Clean up the script string if needed
script = script.strip()
if not script.startswith("[") or not script.endswith("]"):
print("Invalid transcript format. Expected a list of tuples.")
return None
try:
transcript_list = ast.literal_eval(script)
if not isinstance(transcript_list, list):
raise ValueError("Transcript is not a list")
all_audio_segments = []
# Process each dialogue entry
for i, entry in enumerate(transcript_list):
if not isinstance(entry, tuple) or len(entry) != 2:
print(f"Skipping invalid entry {i}: {entry}")
continue
speaker, dialogue = entry
chosen_voice = voice_map.get(speaker, "af_heart")
print(f"Generating audio for {speaker} with voice '{chosen_voice}'...")
pipeline = KPipeline(lang_code="a")
generator = pipeline(dialogue, voice=chosen_voice)
segment_audio = []
for j, (gs, ps, audio) in enumerate(generator):
# print(f"{speaker} - Segment {j}: Global Step = {gs}, Partial Step = {ps}")
segment_audio.append(audio)
if segment_audio:
segment_full = np.concatenate(segment_audio, axis=0)
all_audio_segments.append(segment_full)
if not all_audio_segments:
print("No audio segments were generated.")
return None
# Add a pause between segments
sample_rate = 24000
pause = np.zeros(sample_rate, dtype=np.float32)
final_audio = all_audio_segments[0]
for seg in all_audio_segments[1:]:
final_audio = np.concatenate((final_audio, pause, seg), axis=0)
sf.write(output_file, final_audio, sample_rate)
print(f"Saved final audio as {output_file}")
return output_file
except Exception as e:
print(f"Error processing transcript: {e}")
return None
def process_pdf(pdf_file, speaker1_voice, speaker2_voice, provider, api_key, openrouter_base=None):
"""Process the uploaded PDF file and generate audio"""
try:
# Set API configuration based on provider
if provider == "openai":
os.environ["OPENAI_API_KEY"] = api_key
os.environ["OPENROUTER_API_BASE"] = "https://api.openai.com/v1"
else:
os.environ["OPENAI_API_KEY"] = api_key
os.environ["OPENROUTER_API_BASE"] = openrouter_base or "https://openrouter.ai/api/v1"
# Check if we received a valid file
if pdf_file is None:
return "No file uploaded", None
# Create a temporary file with .pdf extension
with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
# For Gradio uploads, we need to copy the file
shutil.copy2(pdf_file.name, tmp.name)
tmp_path = tmp.name
print(f"Uploaded PDF saved at {tmp_path}")
# Generate transcript using your existing function
transcript, transcript_path = generate_podcast_script(tmp_path, provider=provider)
if transcript is None:
return "Error generating transcript", None
# Define an output file path for the generated audio
audio_output_path = os.path.join(
os.path.dirname(tmp_path),
f"audio_{os.path.basename(tmp_path).replace('.pdf', '.wav')}"
)
result = generate_audio_from_script_with_voices(
transcript,
speaker1_voice,
speaker2_voice,
output_file=audio_output_path
)
if result is None:
return "Error generating audio", None
return "Process complete!", result
except Exception as e:
print(f"Error in process_pdf: {str(e)}")
return f"Error processing file: {str(e)}", None
def create_gradio_app():
# Add CSS for better styling
css = """
.gradio-container {max-width: 900px !important}
"""
with gr.Blocks(css=css, theme=gr.themes.Soft()) as app:
gr.Markdown(
"""
# 📚 NotebookLM-Kokoro TTS App
Upload a PDF, choose voices, and generate conversational audio using Kokoro TTS.
"""
)
with gr.Row():
with gr.Column(scale=2):
pdf_input = gr.File(
label="Upload PDF Document",
file_types=[".pdf"],
type="filepath"
)
with gr.Row():
speaker1_voice = gr.Dropdown(
choices=["af_heart", "af_bella", "hf_beta"],
value="af_heart",
label="Speaker 1 Voice"
)
speaker2_voice = gr.Dropdown(
choices=["af_nicole", "af_heart", "bf_emma"],
value="af_nicole",
label="Speaker 2 Voice"
)
provider = gr.Radio(
choices=["openai", "openrouter"],
value="openrouter",
label="API Provider"
)
with gr.Group():
provider = gr.Radio(
choices=["openai", "openrouter"],
value="openrouter",
label="API Provider"
)
api_key = gr.Textbox(
label="API Key",
placeholder="Enter your API key here...",
type="password",
class_name="api-key-input"
)
openrouter_base = gr.Textbox(
label="OpenRouter Base URL (optional)",
placeholder="https://openrouter.ai/api/v1",
visible=False
)
# Show/hide OpenRouter base URL based on provider selection
def toggle_openrouter_base(provider_choice):
return gr.update(visible=provider_choice == "openrouter")
provider.change(
fn=toggle_openrouter_base,
inputs=[provider],
outputs=[openrouter_base]
)
submit_btn = gr.Button("🎙️ Generate Audio", variant="primary")
with gr.Column(scale=2):
status_output = gr.Textbox(
label="Status",
placeholder="Processing status will appear here..."
)
audio_output = gr.Audio(
label="Generated Audio",
type="filepath"
)
# Examples section
gr.Examples(
examples=[
["sample.pdf", "af_heart", "af_nicole", "openrouter"],
],
inputs=[pdf_input, speaker1_voice, speaker2_voice, provider],
outputs=[status_output, audio_output],
fn=process_pdf,
cache_examples=True,
)
submit_btn.click(
fn=process_pdf,
inputs=[pdf_input, speaker1_voice, speaker2_voice, provider],
outputs=[status_output, audio_output],
api_name="generate"
)
gr.Markdown(
"""
### 📝 Notes
- Make sure your PDF is readable and contains text (not scanned images)
- Processing large PDFs may take a few minutes
- You need a valid OpenAI/OpenRouter API key set as environment variable
"""
)
return app
if __name__ == "__main__":
demo = create_gradio_app()
demo.queue(concurrency_count=1).launch(
server_name="0.0.0.0",
server_port=7860,
share=True,
debug=True
) |