接口说明

接入必读

请先查看接入必读了解具体接入方式，再参考此文档完成接入。

功能介绍

MemeSong为用户提供文本转歌曲的歌唱合成能力，该API目前支持非流式、中文合成。

• 输入：待合成的有效中文可读文本
• 输出：歌曲音频二进制数据

接口说明

• 当前支持通过 HTTP 协议在线调用

• 请求内容包括：payload字段为将请求参数序列化后的json文本

• 使用备注：

公共参数

参考详细说明功能调用-通用协议。

配置参数

payload配置参数为json字符串格式

响应格式

HTTP响应Content-Type: application/json

• 歌唱合成结果在data字段中：返回音频的二进制数据，标准base64编码

参考示例

调用方式为：POST /api/v1/invoke

Golang

// Code sample:
// use http client to invoke SAMI HTTP Service
package main

import (
	"bytes"
	"encoding/json"
	"fmt"
	"io/ioutil"
	"log"
	"net/http"
	"time"
)

type InvokeResponse struct {
	StatusCode int32   `form:"status_code,required" json:"status_code,required" query:"status_code,required"`
	StatusText string  `form:"status_text,required" json:"status_text,required" query:"status_text,required"`
	TaskId     string  `form:"task_id,required" json:"task_id,required" query:"task_id,required"`
	Namespace  string  `form:"namespace,required" json:"namespace,required" query:"namespace,required"`
	Payload    *string `form:"payload,omitempty" json:"payload,omitempty" query:"payload,omitempty"`
	Data       []byte  `form:"data,omitempty" json:"data,omitempty" query:"data,omitempty"`
	State      *string `form:"state,omitempty" json:"state,omitempty" query:"state,omitempty"`
}

const (
	domain = "https://sami.bytedance.com"

	// auth token
	appkey = "your_appkey"

	// SAMI method
	version   = "v4"
	namespace = "MemeSong"

	// dump output
	dataOutputFile    = "output.wav"
	payloadOutputFile = "output.json"
	isDump            = true
)

func main() {
	// Get token
	token := "your_token"

	// Construct HTTP request
	//   1. Read local audio file and construct request payload
	//   2. Set HTTP json body
	//   3. Do HTTP POST request
	speaker := "zh_female_sing_qingxin"
	songName := "xinnianhao_f"
	text := "老师下课后内容没讲完就拖长了课堂 昨天晚上睡觉前我调好了闹钟 过去我是一个小懒猪 每天早上都会赖床"
	body := fmt.Sprintf(
		`{"payload":"{\"speaker\":\"%v\",\"song_name\":\"%v\",\"text\":\"%v\",\"audio_config\": {\"format\":\"wav\"}}"}`,
		speaker, songName, text,
	)
	urlPath := fmt.Sprintf(
		"%v/api/v1/invoke?version=%v&token=%v&appkey=%v&namespace=%v",
		domain, version, token, appkey, namespace,
	)
	log.Printf("invoke request: %v", urlPath)

	// HTTP POST request
	start := time.Now()
	resp, err := http.Post(urlPath, "application/json", bytes.NewBuffer([]byte(body)))
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()

	// Parse HTTP response
	ret, err := ioutil.ReadAll(resp.Body)
	if err != nil || resp.StatusCode != http.StatusOK {
		panic(string(ret))
	}
	log.Printf("http invoke: cost=%vms", time.Since(start).Milliseconds())

	// parse SAMI response
	samiResp := InvokeResponse{}
	payloadStr := ""
	if err = json.Unmarshal(ret, &samiResp); err != nil {
		log.Println("parse response failed", string(ret), err)
		panic(err)
	}
	if samiResp.Payload != nil {
		payloadStr = *samiResp.Payload
	}
	log.Printf("response task_id=%v, payload=%v, data=[%d]byte", samiResp.TaskId, payloadStr, len(samiResp.Data))
	if isDump && samiResp.Payload != nil {
		_ = ioutil.WriteFile(payloadOutputFile, []byte(*samiResp.Payload), 0644)
	}
	if isDump && len(samiResp.Data) > 0 {
		_ = ioutil.WriteFile(dataOutputFile, samiResp.Data, 0644)
	}
}

Python

import base64
import json

import requests
import sys

# Construct HTTP request
payload = json.dumps({
    "text": "老师下课后内容没讲完就拖长了课堂 昨天晚上睡觉前我调好了闹钟 过去我是一个小懒猪 每天早上都会赖床",
    "speaker": "zh_female_sing_qingxin",
    "song_name": "xinnianhao_f",
    "audio_config": {
        "format": "wav",
        "sample_rate": 24000,
        "speech_rate": 0,
    },
})
req = {
    "appkey": "your_appkey",
    "token": "your_token",
    "namespace": "MemeSong",
    "payload": payload,
}

if __name__ == "__main__":
    # HTTP POST request
    resp = requests.post("https://sami.bytedance.com/api/v1/invoke", json=req)

    # Parse HTTP SAMI response
    try:
        sami_resp = resp.json()
        if resp.status_code != 200:
            print(sami_resp)
            sys.exit(1)
    except:
        print(resp)
        sys.exit(1)

    print("response task_id=%s, status_code=%d, status_text=%s" % (
        sami_resp["task_id"], sami_resp["status_code"], sami_resp["status_text"]))

    # Save audio data into file
    if sami_resp["status_code"] == 20000000 and len(sami_resp["data"]) > 0:
        audio_data = base64.b64decode(sami_resp["data"])
        print("audio_data_len: ", len(audio_data))
        with open("audio.wav", "wb") as f:
            f.write(audio_data)

常见问题

• 通用FAQ

• 通用状态码

附录

发音人列表

歌曲列表