文本转语音 AI 2026：OpenAI TTS、ElevenLabs 与语音克隆

现代 AI TTS 能够生成与人类语音难以区分的自然语音。

API 对比 2026

服务质量语言语音克隆成本

OpenAI TTS高57否$15/百万字符 ElevenLabs最高29是$5/万字符 Cartesia高15是$5/百万字符 Kokoro (本地)良好英/日否免费

OpenAI TTS API

python
from openai import OpenAI
from pathlib import Path
client = OpenAI()
生成语音
speech = client.audio.speech.create(
    model='tts-1-hd',  # tts-1 (快速) 或 tts-1-hd (更高质量)
    voice='alloy',  # alloy, echo, fable, onyx, nova, shimmer
    input='欢迎来到 AI Skill Navigator。本教程涵盖文本转语音 API。'
)
speech.stream_to_file('output.mp3')
实时流式传输到扬声器
import pyaudio
p = pyaudio.PyAudio()
stream = p.open(format=8, channels=1, rate=24000, output=True)
with client.audio.speech.with_streaming_response.create(
    model='tts-1',
    voice='nova',
    input='这段文本在生成的同时实时流式传输到扬声器。'
) as response:
    for chunk in response.iter_bytes(1024):
        stream.write(chunk)
stream.close()
p.terminate()

ElevenLabs API (最佳质量)

python
from elevenlabs import ElevenLabs, save
client = ElevenLabs(api_key='your-api-key')
文本转语音
audio = client.text_to_speech.convert(
    voice_id='21m00Tcm4TlvDq8ikWAM',  # 语音 ID (Rachel)
    text='你好！这是 ElevenLabs 生成的 AI 语音。',
    model_id='eleven_multilingual_v2',
    voice_settings={
        'stability': 0.5,
        'similarity_boost': 0.8,
        'style': 0.3,
        'use_speaker_boost': True
    }
)
save(audio, 'elevenlabs_output.mp3')
克隆语音
voice = client.voices.clone(
    name='我的自定义语音',
    description='专业旁白语音',
    files=['sample1.mp3', 'sample2.mp3', 'sample3.mp3'],  # 1-30 个样本
)
print(f'新语音 ID: {voice.voice_id}')
使用克隆的语音
audio = client.text_to_speech.convert(
    voice_id=voice.voice_id,
    text='这是使用我克隆的语音生成的！'
)

用于聊天机器人的流式 TTS

python
import asyncio
async def speak_stream(text: str):
    """在生成时流式传输 TTS 音频，实现低延迟的聊天机器人响应。"""
    async with client.audio.speech.with_streaming_response.create(
        model='tts-1',
        voice='alloy',
        input=text
    ) as response:
        async for chunk in response.aiter_bytes(1024):
            yield chunk  # 将块传输到音频播放器
在 FastAPI 端点中：
from fastapi import FastAPI
from fastapi.responses import StreamingResponse
app = FastAPI()
@app.post('/speak')
async def speak_endpoint(text: str):
    async def generate():
        async for chunk in speak_stream(text):
            yield chunk
    return StreamingResponse(generate(), media_type='audio/mpeg')

构建语音 AI 助手

python
class VoiceAssistant:
    def __init__(self):
        self.client = OpenAI()
        self.history = []
    
    def listen(self) -> str:
        # 使用 pyaudio 录制音频
        # ...录制代码...
        with open('/tmp/recording.wav', 'rb') as f:
            return self.client.audio.transcriptions.create(model='whisper-1', file=f).text
    
    def think(self, user_input: str) -> str:
        self.history.append({'role': 'user', 'content': user_input})
        r = self.client.chat.completions.create(
            model='gpt-4o',
            messages=[{'role': 'system', 'content': '你是一个有用的语音助手。'}] + self.history
        )
        response = r.choices[0].message.content
        self.history.append({'role': 'assistant', 'content': response})
        return response
    
    def speak(self, text: str):
        speech = self.client.audio.speech.create(model='tts-1', voice='nova', input=text)
        speech.stream_to_file('/tmp/response.mp3')
        import subprocess
        subprocess.run(['afplay', '/tmp/response.mp3'])  # macOS
    
    def run(self):
        while True:
            print('正在聆听...')
            text = self.listen()
            print(f'你: {text}')
            response = self.think(text)
            print(f'助手: {response}')
            self.speak(response)

结论

2026 年的 AI TTS 能够为应用提供自然语音。OpenAI TTS 适用于生产 API，ElevenLabs 适用于语音克隆和最高质量，Kokoro 本地运行可实现免费推理。