Chat Completions での音声

Chat Completions エンドポイントを使ったテキストベースの LLM アプリケーションに、音声機能を追加したい場合があります。たとえば、テキスト入力に対応したチャットアプリケーションでは、modalities 配列に audio を含め、gpt-audio-1.5 などの音声モデルを使用することで、音声入出力を追加できます。

Responses API のドキュメントでは現在、テキストと画像を入力し、テキストを出力する方法を説明しています。この音声チャットの構成では、音声対応モデルと Chat Completions を使用してください。

プロンプトに対する人間らしい音声応答の生成

JavaScript

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
28import { writeFileSync } from "node:fs";
import OpenAI from "openai";

const openai = new OpenAI();

// Generate an audio response to the given prompt
const response = await openai.chat.completions.create({
  model: "gpt-audio-1.5",
  modalities: ["text", "audio"],
  audio: { voice: "alloy", format: "wav" },
  messages: [
    {
      role: "user",
      content: "Is a golden retriever a good family dog?",
    },
  ],
  store: true,
});

// Inspect returned data
console.log(response.choices[0]);

// Write audio data to a file
writeFileSync(
  "dog.wav",
  Buffer.from(response.choices[0].message.audio.data, "base64"),
  { encoding: "utf-8" }
);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17import base64
from openai import OpenAI

client = OpenAI()

completion = client.chat.completions.create(
    model="gpt-audio-1.5",
    modalities=["text", "audio"],
    audio={"voice": "alloy", "format": "wav"},
    messages=[{"role": "user", "content": "Is a golden retriever a good family dog?"}],
)

print(completion.choices[0])

wav_bytes = base64.b64decode(completion.choices[0].message.audio.data)
with open("dog.wav", "wb") as f:
    f.write(wav_bytes)

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
34package main

import (
	"context"
	"encoding/base64"
	"fmt"
	"os"

	"github.com/openai/openai-go/v3"
)

func main() {
	client := openai.NewClient()
	response, err := client.Chat.Completions.New(context.Background(), openai.ChatCompletionNewParams{
		Model:      "gpt-audio-1.5",
		Modalities: []string{"text", "audio"},
		Audio: openai.ChatCompletionAudioParam{
			Voice:  openai.ChatCompletionAudioParamVoiceUnion{OfString: openai.String("alloy")},
			Format: openai.ChatCompletionAudioParamFormatWAV,
		},
		Messages: []openai.ChatCompletionMessageParamUnion{openai.UserMessage("Is a golden retriever a good family dog?")},
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(response.Choices[0])
	audio, err := base64.StdEncoding.DecodeString(response.Choices[0].Message.Audio.Data)
	if err != nil {
		panic(err)
	}
	if err := os.WriteFile("dog.wav", audio, 0o600); err != nil {
		panic(err)
	}
}

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
28import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.chat.completions.ChatCompletionAudioParam;
import com.openai.models.chat.completions.ChatCompletionCreateParams;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.util.Base64;

ChatCompletionCreateParams params =
    ChatCompletionCreateParams.builder()
        .model("gpt-audio-1.5")
        .addUserMessage("Is a golden retriever a good family dog?")
        .addModality(ChatCompletionCreateParams.Modality.TEXT)
        .addModality(ChatCompletionCreateParams.Modality.AUDIO)
        .audio(
            ChatCompletionAudioParam.builder()
                .voice("alloy")
                .format(ChatCompletionAudioParam.Format.WAV)
                .build())
        .store(true)
        .build();

var message = client.chat().completions().create(params).choices().get(0).message();
var audio =
    message.audio().orElseThrow(() -> new IllegalStateException("No audio output returned"));
Files.write(Path.of("dog.wav"), Base64.getDecoder().decode(audio.data()));
message.content().ifPresent(System.out::println);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25using OpenAI.Chat;
#pragma warning disable OPENAI001

string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
ChatClient client = new("gpt-audio-1.5", key);

ChatCompletionOptions options = new()
{
    ResponseModalities = ChatResponseModalities.Text | ChatResponseModalities.Audio,
    AudioOptions = new(ChatOutputAudioVoice.Alloy, ChatOutputAudioFormat.Wav),
    StoredOutputEnabled = true,
};

ChatCompletion completion = await client.CompleteChatAsync(
    [new UserChatMessage("Is a golden retriever a good family dog?")],
    options
);

if (completion.OutputAudio is not ChatOutputAudio audio)
{
    throw new InvalidOperationException("No audio output was returned.");
}

Console.WriteLine(audio.Transcript);
await File.WriteAllBytesAsync("dog.wav", audio.AudioBytes.ToArray());

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22require "base64"
require "openai"

client = OpenAI::Client.new
completion = client.chat.completions.create(
  model: "gpt-audio-1.5",
  messages: [
    {
      role: :user,
      content: "Is a golden retriever a good family dog?"
    }
  ],
  modalities: [:text, :audio],
  audio: {
    voice: :alloy,
    format: :wav
  },
  store: true
)

audio = completion.choices.fetch(0).message.audio or raise "No audio returned"
File.binwrite("dog.wav", Base64.strict_decode64(audio.data))

1
2
3
4
5
6
7
8
9
10
11
12
13
14curl "https://api.openai.com/v1/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OPENAI_API_KEY" \
    -d '{
      "model": "gpt-audio-1.5",
      "modalities": ["text", "audio"],
      "audio": { "voice": "alloy", "format": "wav" },
      "messages": [
        {
          "role": "user",
          "content": "Is a golden retriever a good family dog?"
        }
      ]
    }'

モデルへのプロンプトとしての音声入力の使用

JavaScript

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
29import OpenAI from "openai";
const openai = new OpenAI();

// Fetch an audio file and convert it to a base64 string
const url = "https://cdn.openai.com/API/docs/audio/alloy.wav";
const audioResponse = await fetch(url);
const buffer = await audioResponse.arrayBuffer();
const base64str = Buffer.from(buffer).toString("base64");

const response = await openai.chat.completions.create({
  model: "gpt-audio-1.5",
  modalities: ["text", "audio"],
  audio: { voice: "alloy", format: "wav" },
  messages: [
    {
      role: "user",
      content: [
        { type: "text", text: "What is in this recording?" },
        {
          type: "input_audio",
          input_audio: { data: base64str, format: "wav" },
        },
      ],
    },
  ],
  store: true,
});

console.log(response.choices[0]);

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
32import base64
import requests
from openai import OpenAI

client = OpenAI()

# Fetch the audio file and convert it to a base64 encoded string
url = "https://cdn.openai.com/API/docs/audio/alloy.wav"
response = requests.get(url)
response.raise_for_status()
wav_data = response.content
encoded_string = base64.b64encode(wav_data).decode("utf-8")

completion = client.chat.completions.create(
    model="gpt-audio-1.5",
    modalities=["text", "audio"],
    audio={"voice": "alloy", "format": "wav"},
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "What is in this recording?"},
                {
                    "type": "input_audio",
                    "input_audio": {"data": encoded_string, "format": "wav"},
                },
            ],
        },
    ],
)

print(completion.choices[0].message)

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
37package main

import (
	"context"
	"encoding/base64"
	"fmt"
	"os"

	"github.com/openai/openai-go/v3"
)

func main() {
	audio, err := os.ReadFile("fixtures/audio.wav")
	if err != nil {
		panic(err)
	}
	client := openai.NewClient()
	response, err := client.Chat.Completions.New(context.Background(), openai.ChatCompletionNewParams{
		Model:      "gpt-audio-1.5",
		Modalities: []string{"text", "audio"},
		Audio: openai.ChatCompletionAudioParam{
			Voice:  openai.ChatCompletionAudioParamVoiceUnion{OfString: openai.String("alloy")},
			Format: openai.ChatCompletionAudioParamFormatWAV,
		},
		Messages: []openai.ChatCompletionMessageParamUnion{openai.UserMessage([]openai.ChatCompletionContentPartUnionParam{
			openai.TextContentPart("What is in this recording?"),
			openai.InputAudioContentPart(openai.ChatCompletionContentPartInputAudioInputAudioParam{
				Data:   base64.StdEncoding.EncodeToString(audio),
				Format: "wav",
			}),
		})},
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(response.Choices[0])
}

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
54import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.chat.completions.ChatCompletionAudioParam;
import com.openai.models.chat.completions.ChatCompletionContentPart;
import com.openai.models.chat.completions.ChatCompletionContentPartInputAudio;
import com.openai.models.chat.completions.ChatCompletionContentPartText;
import com.openai.models.chat.completions.ChatCompletionCreateParams;
import com.openai.models.chat.completions.ChatCompletionUserMessageParam;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.util.Base64;
import java.util.List;

String encodedAudio =
    Base64.getEncoder()
        .encodeToString(
            Files.readAllBytes(Path.of(System.getenv("OPENAI_EXAMPLE_AUDIO_PATH"))));

ChatCompletionCreateParams params =
    ChatCompletionCreateParams.builder()
        .model("gpt-audio-1.5")
        .addMessage(
            ChatCompletionUserMessageParam.builder()
                .contentOfArrayOfContentParts(
                    List.of(
                        ChatCompletionContentPart.ofText(
                            ChatCompletionContentPartText.builder()
                                .text("What is in this recording?")
                                .build()),
                        ChatCompletionContentPart.ofInputAudio(
                            ChatCompletionContentPartInputAudio.builder()
                                .inputAudio(
                                    ChatCompletionContentPartInputAudio.InputAudio.builder()
                                        .data(encodedAudio)
                                        .format(
                                            ChatCompletionContentPartInputAudio.InputAudio
                                                .Format.WAV)
                                        .build())
                                .build())))
                .build())
        .addModality(ChatCompletionCreateParams.Modality.TEXT)
        .addModality(ChatCompletionCreateParams.Modality.AUDIO)
        .audio(
            ChatCompletionAudioParam.builder()
                .voice("alloy")
                .format(ChatCompletionAudioParam.Format.WAV)
                .build())
        .store(true)
        .build();

client.chat().completions().create(params).choices().stream()
    .flatMap(choice -> choice.message().content().stream())
    .forEach(System.out::println);

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
33using OpenAI.Chat;
#pragma warning disable OPENAI001

string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
ChatClient client = new("gpt-audio-1.5", key);

BinaryData audio = BinaryData.FromBytes(
    await File.ReadAllBytesAsync("audio.wav")
);
UserChatMessage message = new(
    [
        ChatMessageContentPart.CreateTextPart("What is in this recording?"),
        ChatMessageContentPart.CreateInputAudioPart(
            audio,
            ChatInputAudioFormat.Wav
        ),
    ]
);
ChatCompletionOptions options = new()
{
    ResponseModalities = ChatResponseModalities.Text | ChatResponseModalities.Audio,
    AudioOptions = new(ChatOutputAudioVoice.Alloy, ChatOutputAudioFormat.Wav),
    StoredOutputEnabled = true,
};

ChatCompletion completion = await client.CompleteChatAsync([message], options);

if (completion.OutputAudio is not ChatOutputAudio audioOutput)
{
    throw new InvalidOperationException("No audio output was returned.");
}

Console.WriteLine(audioOutput.Transcript);

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
34require "base64"
require "openai"

client = OpenAI::Client.new
audio = Base64.strict_encode64(File.binread("audio.wav"))
completion = client.chat.completions.create(
  model: "gpt-audio-1.5",
  messages: [
    {
      role: :user,
      content: [
        {
          type: :text,
          text: "What is in this recording?"
        },
        {
          type: :input_audio,
          input_audio: {
            data: audio,
            format: :wav
          }
        }
      ]
    }
  ],
  modalities: [:text, :audio],
  audio: {
    voice: :alloy,
    format: :wav
  },
  store: true
)

puts(completion.choices.fetch(0).message.content)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23curl "https://api.openai.com/v1/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OPENAI_API_KEY" \
    -d '{
      "model": "gpt-audio-1.5",
      "modalities": ["text", "audio"],
      "audio": { "voice": "alloy", "format": "wav" },
      "messages": [
        {
          "role": "user",
          "content": [
            { "type": "text", "text": "What is in this recording?" },
            { 
              "type": "input_audio", 
              "input_audio": { 
                "data": "<base64 bytes here>", 
                "format": "wav" 
              }
            }
          ]
        }
      ]
    }'

おすすめ

はじめに

基本コンセプト

SDK と CLI

リソース

レガシー API

モデルの選び方

テキストとコード

プロンプト

推論

画像と動画

リアルタイムと音声

特化型モデル

Agents API

Agents SDK

ChatKit

検索と取得

ツールとデータの接続

ツールを使ったワークフローの構築

コンピューターとコード

メディア

GPT-Live

Realtime API

音声を活用した開発

接続

音声処理

本番運用の開始

性能と品質

コストとスループット

安全性とガバナンス

インフラとアクセス

オペレーション

基本コンセプト

Plan

Build

Test and publish

Conversion specs

ガイド

リソース

はじめに

ガイド

File Upload

API

Measurement

Advertiser API

API リファレンス

はじめに

基礎

詳しく見る

利用可能な環境

リリース

ワークフロー

できること

リファレンス

カスタマイズ

設定ファイル

エージェントの構成

ChatGPT と Codex の機能拡張

Linux

Windows

開発ワークフロー

拡張と自動化

環境

Codex を使った開発

サードパーティ連携

リファレンス

権限

Codex Security

サイバーセーフティ

はじめに

ChatGPT Work

ID と認証

ワークスペースのアクセス、ポリシー、モデル

プラグインとコネクタの制御

利用状況、ガバナンス、コンプライアンス

展開とモデルプロバイダー

コミュニティ

ブログ

コミュニティ

ブログ