语音合成

import OpenAI from "openai";
import fs from "fs";
import path from "path";

const STEP_API_MODEL = "step-tts-mini";

const openai = new OpenAI({
    apiKey: process.env.STEPFUN_API_KEY,
    baseURL: "https://api.stepfun.com/v1"
});

async function main() {
    const speechFile = path.resolve("./speech.mp3");
    const mp3 = await openai.audio.speech.create({
        model: STEP_API_MODEL,
        voice: "cixingnansheng",
        input: "智能阶跃，十倍每个人的可能.",
        extra_body: {
            volume: 2.0, // volume 在拓展参数里
            // 文本归一化：standard / enhanced，默认 standard
            text_normalization: "standard",
            // voice_label 的 language / emotion / style 同一时间只能设其中一个
            voice_label: {
                emotion: "高兴"
            },
            pronunciation_map: {
                tone: [
                    "阿胶/e1胶",
                    "扁舟/偏舟",
                    "LOL/laugh out loudly"
                ]
            }
        }
    });
    console.log(speechFile);
    const buffer = Buffer.from(await mp3.arrayBuffer());
    await fs.promises.writeFile(speechFile, buffer);
}

main();

curl --location 'https://api.stepfun.com/v1/audio/speech' \
  --header 'Content-Type: application/json' \
  --header "Authorization: Bearer $STEPFUN_API_KEY" \
  --data '{
    "model": "stepaudio-2.5-tts",
    "input": "2025 年第 4 季度营收同比增长 18.6%，约合 12.3 亿元。",
    "voice": "cixingnansheng",
    "text_normalization": "enhanced"
  }' \
  --output "step.mp3"

import os
from pathlib import Path
from openai import OpenAI

speech_file_path = Path(__file__).parent / "step-tts.mp3"

client = OpenAI(
    api_key=os.environ["STEPFUN_API_KEY"],
    base_url="https://api.stepfun.com/v1"
)
with client.audio.speech.with_streaming_response.create(
    model="stepaudio-2.5-tts",
    voice="cixingnansheng",
    # 长度需 <=1000 字符；括号 () 内作为指令处理不发音，需发音文本请勿加括号
    input="（冷笑）你以为我们阶跃星辰北京公司的技术是开玩笑的吗！",
    extra_body={
        "volume": 1.0,  # volume 在拓展参数里
        # 文本归一化策略：standard / enhanced，默认 standard；enhanced 适合播报等更看重合成效果的场景
        "text_normalization": "enhanced",
        # 全局情绪指令，最长 200 字符
        "instruction": "语气极其愤怒，压迫感强，语速偏快",
        "pronunciation_map": {
            "tone": [
                "阿胶/e1胶",
                "扁舟/偏舟",
                "LOL/laugh out loudly"
            ]
        }
    }
) as response:
    response.stream_to_file(speech_file_path)

import os
import requests

url = "https://api.stepfun.com/v1/audio/speech"
headers = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {os.environ['STEPFUN_API_KEY']}"
}
payload = {
    "model": "step-tts-mini",
    "input": "今天天气真好。我们出去玩吧。",
    "voice": "cixingnansheng",
    "response_format": "mp3",
    "sample_rate": 24000,
    # 开启字幕；非流式需同时设置 return_url=true，字幕随文件 URL 返回
    "return_url": True,
    "timestamp": True
}

result = requests.post(url, headers=headers, json=payload).json()

# 音频文件下载地址
print("音频地址:", result["data"]["url"])

# 逐句、逐词时间戳
for subtitle in result["data"]["subtitles"]:
    print(f"\n切句: {subtitle['text']}")
    for item in subtitle["items"]:
        print(f"  {item['text']}: {item['start_time']}ms - {item['end_time']}ms")

import os
import json
import requests

url = "https://api.stepfun.com/v1/audio/speech"
headers = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {os.environ['STEPFUN_API_KEY']}"
}
payload = {
    "model": "step-tts-mini",
    "input": "今天天气真好。我们出去玩吧。",
    "voice": "cixingnansheng",
    "response_format": "mp3",
    "sample_rate": 24000,
    # 开启字幕；流式需设置 stream_format="sse"，字幕以 response.subtitle 事件下发
    "stream_format": "sse",
    "timestamp": True
}

resp = requests.post(url, headers=headers, json=payload, stream=True)

audio_chunks = []
for line in resp.iter_lines():
    if not line:
        continue
    line = line.decode("utf-8")
    if not line.startswith("data: "):
        continue
    data_str = line[6:]
    if data_str == "[DONE]":
        break
    event = json.loads(data_str)

    if event["type"] == "speech.audio.delta":
        # 音频分片：BASE64，拼接后即完整音频
        audio_chunks.append(event["audio"])
    elif event["type"] == "response.subtitle":
        # 字幕事件：该切句的逐词时间戳
        data = event["data"]
        print(f"字幕: {data['text']}")
        for item in data["items"]:
            print(f"  {item['text']}: {item['start_time']}ms - {item['end_time']}ms")
    elif event["type"] == "speech.audio.done":
        print("音频生成完成")