File size: 2,948 Bytes
c967100 |
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 |
import sounddevice as sd
import webrtcvad
import numpy as np
from scipy.io.wavfile import write
from faster_whisper import WhisperModel
import time
import os
SAMPLE_RATE = 16000
FRAME_DURATION = 30 # ms
FRAME_SIZE = int(SAMPLE_RATE * FRAME_DURATION / 1000)
SILENCE_THRESHOLD = 0.5 # 秒，连续静音多久算结束
MAX_RECORD_SECONDS = 15 # 最大安全时长
MIN_SPEECH_DURATION = 0.3 # 忽略太短的无效段落
# ✅ 初始化 Whisper 模型（只加载一次）
print("📥 Loading Whisper model...")
model = WhisperModel("small", device="cpu", compute_type="int8")
def record_and_detect(filename="audio.wav"):
vad = webrtcvad.Vad(2)
frames = []
silence_counter = 0
speech_detected = False
max_silence_frames = int(SILENCE_THRESHOLD * 1000 / FRAME_DURATION)
stream = sd.InputStream(samplerate=SAMPLE_RATE, channels=1, dtype='int16', blocksize=FRAME_SIZE)
stream.start()
print("🎙️ 说话开始（说完停顿自动结束）...")
try:
while True:
frame, _ = stream.read(FRAME_SIZE)
pcm = frame.flatten()
pcm_bytes = pcm.tobytes()
is_speech = vad.is_speech(pcm_bytes, SAMPLE_RATE)
frames.append((pcm.copy(), is_speech))
if is_speech:
silence_counter = 0
speech_detected = True
else:
silence_counter += 1
if speech_detected and silence_counter >= max_silence_frames:
print("🛑 停顿检测完成，结束录音")
break
finally:
stream.stop()
stream.close()
# ✅ 剪掉尾部静音帧
cut_index = len(frames)
for i in range(len(frames) - 1, -1, -1):
if frames[i][1]: # 是语音
cut_index = i + 1
break
trimmed_audio = np.concatenate([frames[i][0] for i in range(cut_index)])
duration = len(trimmed_audio) / SAMPLE_RATE
if duration < MIN_SPEECH_DURATION:
print("⚠️ 忽略无效短录音")
return None
write(filename, SAMPLE_RATE, trimmed_audio.astype(np.int16))
print(f"💾 已保存音频：{filename} (长度: {duration:.2f}s)")
return filename
def transcribe(filename):
print("🔍 开始转录...")
t1 = time.time()
segments, info = model.transcribe(filename, beam_size=3)
t2 = time.time()
print(f"✅ 检测语言: {info.language}")
segment_list = list(segments)
if not segment_list:
print("⚠️ 没识别到语音内容")
else:
print("📄 识别内容：")
for seg in segment_list:
print(f"[{seg.start:.2f}s → {seg.end:.2f}s] {seg.text}")
print(f"⏱️ 转录耗时：{t2 - t1:.2f}s")
if __name__ == "__main__":
while True:
audio_file = record_and_detect()
if audio_file:
transcribe(audio_file)
print("\n✅ 等待下一轮语音输入（Ctrl+C退出）...\n")
|