实时音频转音素实现2D角色唇语同步教程

pip install SpeechRecognition
pip install pyaudio  # 用于麦克风输入

import speech_recognition as sr

def recognize_audio_stream():
    r = sr.Recognizer()
    with sr.Microphone() as source:
        print("请开始说话...")
        r.adjust_for_ambient_noise(source) # 调整环境噪音
        while True:
            try:
                audio = r.listen(source, phrase_time_limit=5) # 监听5秒
                # 可以选择不同的识别器，例如 Google Web Speech API
                # text = r.recognize_google(audio, language="zh-CN")
                # 或者使用本地的 Whisper 模型 (需要额外安装 openai-whisper)
                # text = r.recognize_whisper(audio)

                # 这里我们以 Google Speech Recognition 为例
                text = r.recognize_google(audio) 
                print(f"识别到文本: {text}")
                yield text # 实时返回识别到的文本
            except sr.UnknownValueError:
                print("无法识别音频")
            except sr.RequestError as e:
                print(f"请求失败; {e}")
            except KeyboardInterrupt:
                print("程序终止。")
                break

# 实时获取文本
# for word in recognize_audio_stream():
#     # 在这里处理获取到的单词
#     pass

pip install cmudict

import cmudict

def get_phonemes_from_word(word):
    cmu_dict = cmudict.dict()
    word = word.lower() # CMU Dict通常使用小写
    if word in cmu_dict:
        # 可能会有多个发音，这里取第一个
        return cmu_dict[word][0] 
    else:
        return None

# 示例
word_to_process = "Hello"
phonemes = get_phonemes_from_word(word_to_process)
if phonemes:
    print(f"'{word_to_process}' 的 CMU 音素: {phonemes}")
else:
    print(f"'{word_to_process}' 未在 CMU Dict 中找到。")

word_to_process = "this"
phonemes = get_phonemes_from_word(word_to_process)
if phonemes:
    print(f"'{word_to_process}' 的 CMU 音素: {phonemes}")

pip install ipa2

from ipa2 import CMU_TO_IPA

def convert_cmu_to_ipa(cmu_phonemes):
    ipa_phonemes = []
    for cmu_ph in cmu_phonemes:
        # CMU_TO_IPA是一个字典，直接映射
        # 注意：CMU Dict的音素通常带有数字表示重音，如 'IH0', 'IH1'
        # 在映射前可能需要去除数字，或使用更复杂的映射逻辑
        cmu_ph_clean = cmu_ph.strip("012") # 移除重音数字
        if cmu_ph_clean in CMU_TO_IPA:
            ipa_phonemes.append(CMU_TO_IPA[cmu_ph_clean])
        else:
            ipa_phonemes.append(f"[{cmu_ph_clean}](未知)") # 未知音素
    return "".join(ipa_phonemes)

# 结合之前的示例
word_to_process = "this"
cmu_phonemes = get_phonemes_from_word(word_to_process) # 例如 ['DH', 'IH1', 'S']

if cmu_phonemes:
    print(f"'{word_to_process}' 的 CMU 音素: {cmu_phonemes}")
    ipa_output = convert_cmu_to_ipa(cmu_phonemes)
    print(f"转换为 IPA: {ipa_output}")