Spaces:
Build error
Build error
File size: 3,596 Bytes
4cfc491 7eae664 4cfc491 7eae664 4cfc491 7eae664 4cfc491 7eae664 4cfc491 7eae664 4cfc491 7eae664 4cfc491 | 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 | import re
import numpy as np
from transformers import pipeline
import gradio as gr
# Available voices and their corresponding models
VOICES = {
"Amy (Female)": "microsoft/vits-piper-en-us-amy",
"Joe (Male)": "microsoft/vits-piper-en-us-joe",
"Clara (Female)": "microsoft/vits-piper-en-us-clb",
"Ryan (Male)": "microsoft/vits-piper-en-us-jvs"
}
def parse_segments(text):
"""Parse input text for speaker segments with improved validation"""
pattern = re.compile(r'$$(?P<speaker>[^$$]+?)$$(?P<text>.*?)$$/(?P=speaker)$$', re.DOTALL)
matches = list(pattern.finditer(text))
# Validate speaker names and collect results
valid_segments = []
for match in matches:
speaker = match.group('speaker')
if speaker in VOICES:
valid_segments.append((speaker, match.group('text').strip()))
# Find any invalid segments
if len(matches) < len(text.strip()):
return valid_segments, f"Warning: Found {len(matches)} valid segments, but text contains untagged content or invalid speaker names"
return valid_segments, None
def generate_podcast(input_text):
"""Convert text to podcast with multiple voices"""
try:
segments, warning = parse_segments(input_text)
if not segments:
return (22050, np.zeros(0)), "No valid speaker segments found. Please use the format: [Speaker Name]text[/Speaker Name]"
all_audio = []
current_pipe = None
current_model = ""
for speaker, text in segments:
model_name = VOICES[speaker]
# Load model only when needed
if current_model != model_name:
if current_pipe: del current_pipe
current_pipe = pipeline("text-to-speech", model=model_name)
current_model = model_name
# Generate audio for this segment
output = current_pipe(text)
all_audio.append(output["audio"])
# Combine all audio segments with short pauses
final_audio = np.concatenate([np.concatenate((audio, np.zeros(5000))) for audio in all_audio])
status = "Podcast generated successfully!"
if warning:
status += " " + warning
return (output["sampling_rate"], final_audio), status
except Exception as e:
return (22050, np.zeros(0)), f"Error: {str(e)}"
# Create Gradio interface
def podcast_interface(text):
(sr, audio), status = generate_podcast(text)
return (sr, audio) if audio.size > 0 else gr.update(), status
demo = gr.Interface(
fn=podcast_interface,
inputs=gr.Textbox(
label="Input Text with Speaker Tags",
lines=12,
placeholder="""Example format:
[Amy (Female)]Welcome to our podcast![/Amy (Female)]
[Joe (Male)]Today we're discussing AI innovations.[/Joe (Male)]"""
),
outputs=[
gr.Audio(label="Generated Podcast", type="numpy"),
gr.Textbox(label="Status", value="Ready")
],
examples=[
["""[Amy (Female)]Welcome to our podcast![/Amy (Female)]
[Joe (Male)]Today we're discussing AI innovations.[/Joe (Male)]
[Clara (Female)]This is Clara speaking![/Clara (Female)]"""]
],
title="🎙️ Multi-Voice Podcast Generator",
description="Generate podcasts with multiple free AI voices using Microsoft's Piper TTS models. Use [SpeakerName] tags to assign different voices to different text segments.",
theme="soft",
allow_flagging="never"
)
if __name__ == "__main__":
demo.launch() |