我目前正在做一个关于语言学习的副项目。主要功能包括使用 AI 生成内容以及将文本转换为音频文件。为了存储音频文件我还需要云存储。成本是我的首要考虑因素因为我认为在云平台之间切换不会太困难。最终我选择了 Google Gemini、Google TTS 和 Cloudflare R2。它们提供了 API 文档和示例但我发现有些部分不够完善所以我决定写一篇文章来分享。我使用的是 Go 语言本文只涵盖基本用法。对于 Google Gemini 和 TTS我使用 RESTful API。虽然它们提供了库但我发现使用 RESTful API 比设置库更方便。1. Google Geminipackage api import ( bytes encoding/json fmt io net/http github.com/spf13/viper ) type Part struct { Text string json:text } type Content struct { Parts []Part json:parts } type Candidates struct { Content Content json:content } type PromptResult struct { Candidates []Candidates json:candidates } func Prompt(prompt string) (*PromptResult, error) { // I use viper to manage environment variables, you can replace this with your api key. url : fmt.Sprintf(https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?key%s, viper.Get(GOOGLE_CLOUD_API_KEY)) // In this example, it sends only one prompt but you can send more information data, err : json.Marshal(map[string]interface{}{ contents: []map[string]interface{}{{ parts: []map[string]interface{}{{ text: prompt, }}, }}, }) if err ! nil { return nil, err } req, err : http.NewRequest(POST, url, bytes.NewBuffer(data)) if err ! nil { return nil, err } req.Header.Add(Content-Type, application/json) // Request the API res, err : http.DefaultClient.Do(req) if err ! nil { return nil, err } defer res.Body.Close() resBody, err : io.ReadAll(res.Body) if err ! nil { return nil, err } var promptResult PromptResult // Parse the result err json.Unmarshal(resBody, promptResult) if err ! nil { return nil, err } return promptResult, nil }package main import ( fmt github.com/hsk-kr/tutorial/lib/api github.com/spf13/viper ) func main() { viper.SetConfigFile(.env) viper.ReadInConfig() promptResult, _ : api.Prompt(Generate a short story for kids) fmt.Println(promptResult.Candidates[0].Content.Parts[0].Text) }您可以在以下链接找到有关 API 的更多详细信息https://ai.google.dev/gemini-api/docs。在此示例中发送了一个提示并接收了响应。响应包含的内容不仅仅是文本但为了简单起见我定义的结构体只处理文本并将其打印到控制台。我使用 Viper 来管理环境变量但您可以通过将 API 密钥替换为您自己的来测试代码。2. Google TTSpackage api import ( bytes encoding/base64 encoding/json errors fmt io net/http strings github.com/spf13/viper ) type Voice struct { LanguageCodes []string json:languageCodes Name string json:name SsmlGender Gender json:ssmlGender // MALE or FEMALE NaturalSampleRateHertz int json:naturalSampleRateHertz } type VoiceSelectionParam struct { LanguageCode string json:languageCode Name string json:name SsmlGender Gender json:ssmlGender } type VoicesResponse struct { Voices []Voice json:voices } type AudioEncoding string type Gender string const ( MALE Gender MALE FEMALE Gender FEMALE ) const ( LINEAR16 AudioEncoding LINEAR16 MP3 AudioEncoding MP3 OGG_OPUS AudioEncoding OGG_OPUS MULAW AudioEncoding MULAW ALAW AudioEncoding ALAW ) func convertVoiceToVoiceSelectionParam(voice Voice) (*VoiceSelectionParam, error) { voiceSelectionParam : new(VoiceSelectionParam) if voice.LanguageCodes nil || len(voice.LanguageCodes) 0 { return nil, errors.New(Empty LanguageCodes) } voiceSelectionParam.LanguageCode voice.LanguageCodes[0] voiceSelectionParam.Name voice.Name voiceSelectionParam.SsmlGender voice.SsmlGender return voiceSelectionParam, nil } func GetVoiceList(languageCode string) ([]Voice, error) { url : fmt.Sprintf(https://texttospeech.googleapis.com/v1/voices?languageCode%skey%s, languageCode, viper.GetString(GOOGLE_CLOUD_API_KEY)) req, err : http.NewRequest(GET, url, nil) if err ! nil { return nil, err } req.Header.Add(Content-Type, application/json) res, err : http.DefaultClient.Do(req) if err ! nil { return nil, err } defer res.Body.Close() resBody, err : io.ReadAll(res.Body) if err ! nil { return nil, err } var voicesRes VoicesResponse err json.Unmarshal(resBody, voicesRes) if err ! nil { return nil, err } return voicesRes.Voices, nil } func ConvertTextToAudio(input string, voice Voice) ([]byte, error) { url : fmt.Sprintf(https://texttospeech.googleapis.com/v1/text:synthesize?key%s, viper.GetString(GOOGLE_CLOUD_API_KEY)) voiceSelectionParam, err : convertVoiceToVoiceSelectionParam(voice) if err ! nil { return nil, err } data, err : json.Marshal(map[string]interface{}{ input: map[string]string{text: input}, voice: voiceSelectionParam, audioConfig: map[string]string{audioEncoding: string(OGG_OPUS)}, }) if err ! nil { return nil, err } req, err : http.NewRequest(POST, url, bytes.NewBuffer(data)) if err ! nil { return nil, err } req.Header.Add(Content-Type, application/json) res, err : http.DefaultClient.Do(req) if err ! nil { return nil, err } defer res.Body.Close() body, err : io.ReadAll(res.Body) if err ! nil { return nil, err } var result map[string]interface{} if err : json.Unmarshal(body, result); err ! nil { return nil, err } audioContent, ok : result[audioContent].(string) if !ok { return nil, fmt.Errorf(No audio content found in response) } audioData, err : base64.StdEncoding.DecodeString(audioContent) if err ! nil { return nil, err } return audioData, nil } func getFirstXVoice(voices []Voice, strToFind string, gender Gender) *Voice { for i, v : range voices { if strings.Contains(strings.ToLower(v.Name), strToFind) v.SsmlGender gender { return voices[i] } } return nil } func GetFirstStandardVoice(voices []Voice, gender Gender) *Voice { return getFirstXVoice(voices, standard, gender) } func GetFirstWavenetVoice(voices []Voice, gender Gender) *Voice { return getFirstXVoice(voices, wavenet, gender) } func GetFirstNeuralVoice(voices []Voice, gender Gender) *Voice { return getFirstXVoice(voices, neural, gender) }package main import ( fmt os github.com/hsk-kr/tutorial/lib/api github.com/spf13/viper ) func main() { viper.SetConfigFile(.env) viper.ReadInConfig() voices, _ : api.GetVoiceList(en-US) voice : api.GetFirstWavenetVoice(voices, MALE) bAudio, _ : api.ConvertTextToAudio(By the way, I am using neovim., *voice) os.WriteFile(./audio.opus, bAudio, 0644) }运行程序后您会在同一目录下找到名为 audio.opus 的音频文件。共有三个函数GetVoiceList– 获取 API 支持的语音列表。GetFirstWavenetVoice– 截至 2025 年 2 月 27 日共有三种类型的语音wavenet、standard 和 neural。每种类型包含多个语音但由于这不是我的优先考虑我创建了一个函数来简单地获取给定类型的第一个语音。ConvertTextToAudio– 以文本和语音作为参数返回结果为 []byte。该函数请求 OGG_OPUS 格式的音频文件因为我计划在 Web 环境中使用它。但是您可以使用任何支持的格式。如果您仔细阅读文档可能会发现它缺少一些重要的细节。例如我遇到了 audio_encoding 参数想检查支持哪些格式但没有找到直接链接到该信息的页面。我认为文档可以改进——当我搜索音频编码文档的链接时一无所获只有纯黑色文本。最终我通过在文档顶部手动搜索找到了该文档。3. Cloudflare R2package api import ( bytes context errors fmt time github.com/aws/aws-sdk-go-v2/aws github.com/aws/aws-sdk-go-v2/config github.com/aws/aws-sdk-go-v2/credentials github.com/aws/aws-sdk-go-v2/feature/s3/manager github.com/aws/aws-sdk-go-v2/service/s3 github.com/google/uuid github.com/spf13/viper ) type Storage struct { client *s3.Client uploader *manager.Uploader bucketName string } func (s *Storage) Init() error { cfg, err : config.LoadDefaultConfig(context.TODO(), config.WithCredentialsProvider(credentials.NewStaticCredentialsProvider(viper.GetString(CLOUDFLARE_R2_ACCESS_KEY_ID), viper.GetString(CLOUDFLARE_R2_SECRET_ACCESS_KEY), )), config.WithRegion(auto), ) if err ! nil { return err } s.bucketName viper.GetString(CLOUDFLARE_R2_BUCKET_NAME) s.client s3.NewFromConfig(cfg, func(o *s3.Options) { o.BaseEndpoint aws.String(fmt.Sprintf(https://%s.r2.cloudflarestorage.com, viper.GetString(CLOUDFLARE_R2_ACCOUNT_ID))) }) s.uploader manager.NewUploader(s.client) return nil } func (s *Storage) Put(data []byte) (string, error) { if s.client nil { return , errors.New(client is nil) } objectKey, err : uuid.NewUUID() if err ! nil { return , err } bucket : aws.String(s.bucketName) key : aws.String(objectKey.String()) ctx : context.Background() input : s3.PutObjectInput{ Bucket: bucket, Key: key, Body: bytes.NewReader(data), } output, err : s.uploader.Upload(ctx, input) if err ! nil { return , err } err s3.NewObjectExistsWaiter(s.client).Wait(ctx, s3.HeadObjectInput{ Bucket: bucket, Key: key, }, time.Minute) if err ! nil { return , err } return *output.Key, nil }package main import ( github.com/tutorial/justsayit/lib/api github.com/spf13/viper ) func main() { viper.SetConfigFile(.env) viper.ReadInConfig() voices, _ : api.GetVoiceList(en-US) voice : api.GetFirstWavenetVoice(voices, MALE) bAudio, _ : api.ConvertTextToAudio(By the way, I am using neovim., *voice) storage : new(api.Storage) storage.Init() storage.Put(bAudio) }Cloudflare R2 类似于 Amazon S3因此您可以使用 AWS S3 库来操作其 API。您可以在 AWS 文档中找到更多示例https://docs.aws.amazon.com/code-library/latest/ug/go_2_s3_code_examples.html。由于在将对象上传到云端之前需要创建一些实例我在结构体中定义了必要的方法。运行程序后您应该能看到对象成功上传到 Cloudflare。为了测试音频文件是否能在 Web 环境中正确加载我将对象 URL 放在了标签中——但没有起作用。原来即使我已将文件设置为公开访问仍然需要使用代理来访问它。从安全角度来看这是合理的因为必须在设置中明确允许访问。如果您想临时访问文件进行测试可以启用开发模式并使用开发链接。4. 最终代码package main import ( github.com/tutorial/justsayit/lib/api github.com/spf13/viper ) func main() { viper.SetConfigFile(.env) viper.ReadInConfig() promptResult, _ : api.Prompt(Say something short in German) generatedText : promptResult.Candidates[0].Content.Parts[0].Text voices, _ : api.GetVoiceList(de-DE) voice : api.GetFirstWavenetVoice(voices, MALE) bAudio, _ : api.ConvertTextToAudio(generatedText, *voice) storage : new(api.Storage) storage.Init() storage.Put(bAudio) }这就是最终代码使用 Google Gemini 生成文本将文本转换为音频文件并将音频文件存储在云平台中。原文链接用 Gemini TTS 生成音频故事 - 汇智网