Áudio no Chat Completions

Se você já tem um aplicativo de LLM baseado em texto que usa o endpoint Chat Completions, talvez queira adicionar recursos de áudio. Por exemplo, se seu aplicativo de chat aceita entrada de texto, você pode adicionar entrada e saída de áudio: inclua audio no array modalities e use um modelo de áudio, como o gpt-audio-1.5.

A documentação da API Responses descreve atualmente entradas de texto e imagem com saídas de texto. Para esse padrão de chat com áudio, use Chat Completions com um modelo compatível com áudio.

Crie uma resposta em áudio com voz semelhante à humana a partir de um prompt

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?"
        }
      ]
    }'

Use entradas de áudio como prompts para um modelo

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" 
              }
            }
          ]
        }
      ]
    }'

Sugestões

Primeiros passos

Conceitos fundamentais

SDKs e CLI

Recursos

APIs legadas

Escolha um modelo

Texto e código

Criação de prompts

Raciocínio

Imagens e vídeo

Tempo real e áudio

Modelos especializados

API de Agentes

SDK de Agentes

ChatKit

Pesquisa e recuperação

Conecte ferramentas e dados

Crie fluxos de trabalho com ferramentas

Computador e código

Mídia

GPT-Live

Realtime API

Desenvolva com voz

Conexões

Processamento de áudio

Entre em produção

Desempenho e qualidade

Custo e taxa de processamento

Segurança e governança

Infraestrutura e acesso

Operações

Conceitos fundamentais

Plan

Build

Test and publish

Conversion specs

Guias

Recursos

Primeiros passos

Guias

File Upload

API

Measurement

Advertiser API

Referência da API

Primeiros passos

Fundamentos

Explorar

Disponível em

Versões

Fluxos de trabalho

Capacidades

Referência

Personalização

Arquivo de configuração

Configuração de agentes

Amplie os recursos do ChatGPT e do Codex

Linux

Windows

Fluxos de trabalho de desenvolvimento

Amplie e automatize

Ambientes

Desenvolva com o Codex

Integrações de terceiros

Referência

Permissões

Codex Security

Segurança cibernética

Primeiros passos

ChatGPT Work

Identidade e autenticação

Acesso, política e modelos do workspace

Controles de plug-ins e conectores

Uso, governança e conformidade

Implantação e provedores de modelos

Comunidade

Blog

Comunidade

Blog