Spaces:
Sleeping
Sleeping
File size: 4,621 Bytes
aa8e154 | 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 | """
modules/tts.py — Text-to-Speech (Low-Latency)
Strategy:
1. pyttsx3 — fully offline, zero network latency, engine warmed-up once.
2. gTTS — online fallback if pyttsx3 unavailable.
3. os/wmplayer — last resort Windows fallback.
"""
import threading
import tempfile
import os
# ── pyttsx3 singleton ──────────────────────────────────────────────────────────
_pyttsx3_engine = None
_pyttsx3_lock = threading.Lock()
_pyttsx3_available = False
def _init_pyttsx3():
"""Initialize pyttsx3 engine once at module load (background thread)."""
global _pyttsx3_engine, _pyttsx3_available
try:
import pyttsx3
engine = pyttsx3.init()
# Tune: rate 160 wpm feels natural for an interviewer
engine.setProperty('rate', 160)
engine.setProperty('volume', 1.0)
# Prefer a clear English voice if available
voices = engine.getProperty('voices')
for v in voices:
if 'english' in v.name.lower() or 'zira' in v.name.lower() or 'david' in v.name.lower():
engine.setProperty('voice', v.id)
break
_pyttsx3_engine = engine
_pyttsx3_available = True
print('[TTS] pyttsx3 engine ready (offline, low-latency).')
except Exception as e:
print(f'[TTS] pyttsx3 unavailable ({e}), will fall back to gTTS.')
# Warm-up in background at import time so first speak() has no init overhead
threading.Thread(target=_init_pyttsx3, daemon=True).start()
# ── pygame mixer singleton (for gTTS fallback) ─────────────────────────────────
_pygame_ready = False
def _ensure_pygame():
global _pygame_ready
if _pygame_ready:
return True
try:
import pygame
pygame.mixer.init()
_pygame_ready = True
return True
except Exception:
return False
# ── Public API ─────────────────────────────────────────────────────────────────
def speak(text: str) -> str:
"""
Convert text to speech and play it synchronously.
Returns the path to audio file (if gTTS path was used) or '' for pyttsx3.
Latency profile:
pyttsx3 : ~50-150 ms to first audio sample (fully offline).
gTTS : ~800-2000 ms (network + disk save) — only used as fallback.
"""
if not text or not text.strip():
return ''
# ── Path 1: pyttsx3 (fast offline) ────────────────────────────────────────
with _pyttsx3_lock:
if _pyttsx3_available and _pyttsx3_engine is not None:
try:
_pyttsx3_engine.say(text)
_pyttsx3_engine.runAndWait()
return ''
except Exception as e:
print(f'[TTS] pyttsx3 speak failed ({e}), trying gTTS...')
# ── Path 2: gTTS + pygame ─────────────────────────────────────────────────
tmp_path = ''
try:
from gtts import gTTS
tts = gTTS(text=text, lang='en', slow=False)
tmp_path = tempfile.mktemp(suffix='.mp3')
tts.save(tmp_path)
if _ensure_pygame():
import pygame
pygame.mixer.music.load(tmp_path)
pygame.mixer.music.play()
while pygame.mixer.music.get_busy():
pygame.time.Clock().tick(10)
pygame.mixer.music.unload()
return tmp_path
except Exception as e:
print(f'[TTS] gTTS/pygame failed ({e}), trying wmplayer...')
# ── Path 3: Windows wmplayer ───────────────────────────────────────────────
if tmp_path and os.path.exists(tmp_path):
try:
os.system(f'start /wait wmplayer "{tmp_path}"')
return tmp_path
except Exception:
pass
print(f'[TTS] All playback methods failed. Text: {text[:60]}...')
return tmp_path
def speak_async(text: str) -> threading.Thread:
"""
Fire-and-forget version of speak().
Returns the thread so caller can join() if needed.
"""
t = threading.Thread(target=speak, args=(text,), daemon=True)
t.start()
return t |