Upload 3 files

README.md

@@ -0,0 +1,13 @@
+---
+title: Andrew TTS
+emoji: 🎙️
+colorFrom: yellow
+colorTo: red
+sdk: gradio
+sdk_version: 3.50.2
+app_file: app.py
+pinned: false
+---
+
+# Andrew TTS
+Microsoft Edge TTS with Andrew voice + audio enhancement.

app.py

@@ -0,0 +1,172 @@
+import gradio as gr
+import edge_tts
+import asyncio
+import numpy as np
+import tempfile
+import os
+from scipy import signal
+from scipy.io import wavfile
+from pydub import AudioSegment
+
+VOICE = "en-US-AndrewMultilingualNeural"
+CHUNK_SIZE = 5000
+
+# ── TTS ───────────────────────────────────────────────────────────────────────
+async def generate_chunks(text):
+    chunks = [text[i:i+CHUNK_SIZE] for i in range(0, len(text), CHUNK_SIZE)]
+    parts = []
+    for i, chunk in enumerate(chunks):
+        tmp = tempfile.mktemp(suffix='.mp3')
+        comm = edge_tts.Communicate(chunk, VOICE)
+        await comm.save(tmp)
+        parts.append(tmp)
+    return parts
+
+def merge_audio(parts):
+    combined = AudioSegment.empty()
+    for p in parts:
+        combined += AudioSegment.from_file(p)
+        os.remove(p)
+    return combined
+
+# ── Audio Enhancer ────────────────────────────────────────────────────────────
+def enhance_audio(audio):
+    audio = audio.set_channels(1).set_frame_rate(44100)
+    samples = np.array(audio.get_array_of_samples()).astype(np.float64)
+    samples = samples / (np.max(np.abs(samples)) + 1e-9)
+    sr = 44100
+
+    # Bass boost 100-300Hz +4dB
+    b, a = signal.butter(2, [100/(sr/2), 300/(sr/2)], btype='band')
+    bass = signal.lfilter(b, a, samples)
+    samples = samples + bass * (10**(4/20) - 1)
+
+    # Cut harshness 4-8kHz -3dB
+    b, a = signal.butter(2, [4000/(sr/2), 8000/(sr/2)], btype='band')
+    harsh = signal.lfilter(b, a, samples)
+    samples = samples - harsh * (1 - 10**(-3/20))
+
+    # Proximity effect
+    b, a = signal.butter(1, 200/(sr/2), btype='low')
+    proximity = signal.lfilter(b, a, samples)
+    samples = samples + proximity * 0.15
+
+    # De-esser
+    b, a = signal.butter(2, [6000/(sr/2), min(10000/(sr/2), 0.99)], btype='band')
+    sibilance = signal.lfilter(b, a, samples)
+    sib_env = np.abs(sibilance)
+    threshold = np.percentile(sib_env, 85)
+    mask = sib_env > threshold
+    reduction = np.where(mask, sibilance * 0.4, 0.0)
+    samples = samples - reduction
+
+    # Soft-knee compression
+    thr_lin   = 10 ** (-18/20)
+    knee_lin  = 10 ** (6/20)
+    makeup    = 10 ** (4/20)
+    ratio     = 3.0
+    abs_s     = np.abs(samples)
+    comp      = samples.copy()
+    above     = abs_s > thr_lin * knee_lin
+    in_knee   = (abs_s > thr_lin) & ~above
+    if np.any(above):
+        g = thr_lin * (abs_s[above] / thr_lin) ** (1/ratio)
+        comp[above] = np.sign(samples[above]) * g
+    if np.any(in_knee):
+        kf = (abs_s[in_knee] - thr_lin) / (thr_lin * (knee_lin - 1))
+        kr = 1 + (ratio - 1) * kf ** 2
+        g  = thr_lin * (abs_s[in_knee] / thr_lin) ** (1/kr)
+        comp[in_knee] = np.sign(samples[in_knee]) * g
+    samples = comp * makeup
+
+    # Normalize
+    samples = samples / (np.max(np.abs(samples)) + 1e-9) * 0.95
+    samples_int = (samples * 32767).astype(np.int16)
+
+    enhanced = AudioSegment(
+        samples_int.tobytes(),
+        frame_rate=sr,
+        sample_width=2,
+        channels=1
+    )
+    return enhanced.set_channels(2)
+
+# ── Main function ─────────────────────────────────────────────────────────────
+def generate(text, enhance):
+    if not text.strip():
+        return None, '❌ Please enter some text!'
+
+    try:
+        parts    = asyncio.run(generate_chunks(text))
+        combined = merge_audio(parts)
+
+        if enhance:
+            combined = enhance_audio(combined)
+
+        out_path = tempfile.mktemp(suffix='.mp3')
+        combined.export(out_path, format='mp3', bitrate='192k')
+        return out_path, '✅ Done! Click play or download below.'
+
+    except Exception as e:
+        return None, '❌ Error: ' + str(e)
+
+# ── CSS ───────────────────────────────────────────────────────────────────────
+css = """
+@import url('https://fonts.googleapis.com/css2?family=Syne:wght@700;800&family=DM+Sans:wght@300;400;500&display=swap');
+* { font-family: 'DM Sans', sans-serif; }
+body, .gradio-container { background: #07070f !important; color: #e8e8f5 !important; }
+.gradio-container { max-width: 800px !important; margin: 0 auto !important; }
+h1 { font-family: 'Syne', sans-serif !important; font-weight: 800 !important;
+     font-size: 2.2rem !important;
+     background: linear-gradient(135deg, #ffd700, #ff8c00, #ff4500) !important;
+     -webkit-background-clip: text !important; -webkit-text-fill-color: transparent !important;
+     text-align: center !important; margin-bottom: 0.2rem !important; }
+.subtitle { text-align:center; color:#5a5a7a; font-size:0.88rem; margin-bottom:1.5rem; }
+label { font-family:'Syne',sans-serif !important; font-weight:700 !important;
+        font-size:0.78rem !important; color:#ffd700 !important;
+        letter-spacing:0.06em !important; text-transform:uppercase !important; }
+textarea { background:#0e0e1c !important; border:1px solid #2a2a3a !important;
+    border-radius:10px !important; color:#e8e8f5 !important; padding:12px !important; font-size:0.95rem !important; }
+textarea:focus { border-color:#ffd700 !important; outline:none !important; }
+.block { background:#0d0d1a !important; border:1px solid #1a1a2e !important;
+         border-radius:14px !important; padding:1.2rem !important; }
+.btn-generate { background:linear-gradient(135deg,#b8860b,#ff8c00) !important;
+    border:none !important; border-radius:12px !important; color:white !important;
+    font-family:'Syne',sans-serif !important; font-weight:700 !important;
+    font-size:1.05rem !important; padding:16px !important; width:100% !important; }
+.btn-generate:hover { opacity:0.85 !important; transform:translateY(-1px) !important; }
+.tip { background:#0e0e1c; border-left:3px solid #ffd700; border-radius:0 8px 8px 0;
+       padding:10px 14px; font-size:0.82rem; color:#9ca3af; margin:0.5rem 0; }
+"""
+
+# ── GUI ───────────────────────────────────────────────────────────────────────
+with gr.Blocks(css=css, title='Andrew TTS') as demo:
+    gr.HTML('<h1>🎙️ Andrew TTS</h1>')
+    gr.HTML('<p class="subtitle">Microsoft Andrew voice · Auto enhanced · Download MP3</p>')
+
+    with gr.Row():
+        with gr.Column(scale=2):
+            text_input = gr.Textbox(
+                label='Your Text',
+                placeholder='Paste your novel chapter or any text here...',
+                lines=12
+            )
+            enhance_toggle = gr.Checkbox(
+                label='✨ Enhance audio (warmer voice, de-essed, compressed)',
+                value=True
+            )
+            generate_btn = gr.Button('🎙️ Generate Audio', elem_classes='btn-generate')
+
+        with gr.Column(scale=1):
+            gr.HTML('<div class="tip">📌 Tips:<br><br>• Works best with chapters under 50,000 characters<br><br>• Enhancement adds warmth and depth<br><br>• Audio saves as MP3 192kbps<br><br>• You can close your PC — this runs on the server!</div>')
+            status_box = gr.Textbox(label='Status', interactive=False, lines=2)
+
+    audio_out = gr.Audio(label='Output Audio', type='filepath')
+
+    generate_btn.click(
+        fn=generate,
+        inputs=[text_input, enhance_toggle],
+        outputs=[audio_out, status_box]
+    )
+
+demo.launch()

requirements.txt

@@ -0,0 +1,6 @@
+edge-tts
+pydub==0.25.1
+pyaudioop
+scipy
+numpy
+gradio==3.50.2