Merge pull request #30 from krillinai/feat-aitool

Feat aitool

Author: maranello-o

.gitignore

@@ -7,5 +7,4 @@ bin/
 models/
 uploads/
 app.log
-build/
-models/
+build/

internal/service/subtitle_service.go

@@ -444,19 +444,31 @@ func (s Service) splitAudio(ctx context.Context, stepParam *types.SubtitleTaskSt
 func (s Service) audioToSrt(ctx context.Context, stepParam *types.SubtitleTaskStepParam) error {
 	log.GetLogger().Info("audioToSubtitle.audioToSrt start", zap.Any("taskId", stepParam.TaskId))
 	var (
+		cancel              context.CancelFunc
 		stepNum             = 0
 		parallelControlChan = make(chan struct{}, config.Conf.App.TranslateParallelNum)
-		eg                  errgroup.Group
+		eg                  *errgroup.Group
 		stepNumMu           sync.Mutex
 		err                 error
 	)
+	ctx, cancel = context.WithCancel(ctx)
+	defer cancel()
+	eg, ctx = errgroup.WithContext(ctx)
 	for _, audioFileItem := range stepParam.SmallAudios {
 		parallelControlChan <- struct{}{}
 		audioFile := audioFileItem
 		eg.Go(func() error {
 			defer func() {
 				<-parallelControlChan
+				if r := recover(); r != nil {
+					log.GetLogger().Error("audioToSubtitle.audioToSrt panic recovered", zap.Any("recover", r))
+				}
 			}()
+			select {
+			case <-ctx.Done():
+				return ctx.Err()
+			default:
+			}
 			// 语音转文字
 			var transcriptionData *types.TranscriptionData
 			for i := 0; i < 3; i++ {
@@ -470,6 +482,7 @@ func (s Service) audioToSrt(ctx context.Context, stepParam *types.SubtitleTaskSt
 				}
 			}
 			if err != nil {
+				cancel()
 				log.GetLogger().Error("audioToSubtitle.audioToSrt.Transcription err", zap.Any("stepParam", stepParam), zap.Error(err))
 				return err
 			}
@@ -486,6 +499,7 @@ func (s Service) audioToSrt(ctx context.Context, stepParam *types.SubtitleTaskSt
 			// 拆分字幕并翻译
 			err = s.splitTextAndTranslate(stepParam.TaskId, stepParam.TaskBasePath, stepParam.TargetLanguage, stepParam.EnableModalFilter, audioFile)
 			if err != nil {
+				cancel()
 				log.GetLogger().Error("audioToSubtitle.audioToSrt.splitTextAndTranslate err", zap.Any("stepParam", stepParam), zap.Error(err))
 				return err
 			}
@@ -500,6 +514,7 @@ func (s Service) audioToSrt(ctx context.Context, stepParam *types.SubtitleTaskSt
 			// 生成时间戳
 			err = s.generateTimestamps(stepParam.TaskId, stepParam.TaskBasePath, stepParam.OriginLanguage, stepParam.SubtitleResultType, audioFile)
 			if err != nil {
+				cancel()
 				log.GetLogger().Error("audioToSubtitle.audioToSrt.generateTimestamps err", zap.Any("stepParam", stepParam), zap.Error(err))
 				return err
 			}
@@ -676,7 +691,7 @@ func getSentenceTimestamps(words []types.Word, sentence string, lastTs float64,
 	var srtSt types.SrtSentence
 	var sentenceWordList []string
 	sentenceWords := make([]types.Word, 0)
-	if language == types.LanguageNameEnglish || language == types.LanguageNameGerman { // 处理方式不同
+	if language == types.LanguageNameEnglish || language == types.LanguageNameGerman || language == types.LanguageNameTurkish { // 处理方式不同
 		sentenceWordList = util.SplitSentence(sentence)
 		if len(sentenceWordList) == 0 {
 			return srtSt, sentenceWords, 0, fmt.Errorf("sentence is empty")

pkg/aliyun/asr.go

@@ -7,6 +7,7 @@ import (
 	"github.com/gorilla/websocket"
 	"go.uber.org/zap"
 	"io"
+	"krillin-ai/internal/storage"
 	"krillin-ai/internal/types"
 	"krillin-ai/log"
 	"net/http"
@@ -154,7 +155,7 @@ type Event struct {
 func processAudio(filePath string) (string, error) {
 	dest := strings.ReplaceAll(filePath, filepath.Ext(filePath), "_mono_16K.mp3")
 	cmdArgs := []string{"-i", filePath, "-ac", "1", "-ar", "16000", "-b:a", "192k", dest}
-	cmd := exec.Command("ffmpeg", cmdArgs...)
+	cmd := exec.Command(storage.FfmpegPath, cmdArgs...)
 	output, err := cmd.CombinedOutput()
 	if err != nil {
 		log.GetLogger().Error("处理音频失败", zap.Error(err), zap.String("audio file", filePath), zap.String("output", string(output)))

pkg/aliyun/voice_clone.go

@@ -107,7 +107,7 @@ func (c *VoiceCloneClient) CosyVoiceClone(voicePrefix, audioURL string) (string,
 	}
 	log.GetLogger().Info("CosyVoiceClone请求完毕", zap.String("Response", resp.String()))
 	if res.Message != "SUCCESS" {
-		log.GetLogger().Error("CosyVoiceClone请求响应错误", zap.String("Message", res.Message))
+		log.GetLogger().Error("CosyVoiceClone请求响应错误", zap.String("Request Id", res.RequestId), zap.Int("Code", res.Code), zap.String("Message", res.Message))
 		return "", fmt.Errorf("CosyVoiceClone请求响应错误: %s", res.Message)
 	}
 	return res.VoiceName, nil

pkg/util/subtitle.go

@@ -180,8 +180,8 @@ func ParseSrtNoTsToSrtBlock(srtNoTsFile string) ([]*SrtBlock, error) {
 }
 
 func SplitSentence(sentence string) []string {
-	// 使用正则表达式移除标点符号和特殊字符（除了字母、数字和空格）
-	re := regexp.MustCompile(`[^\w\s']+`)
+	// 使用正则表达式移除标点符号和特殊字符（保留各语言字母、数字和空格）
+	re := regexp.MustCompile(`[^\p{L}\p{N}\s']+`)
 	cleanedSentence := re.ReplaceAllString(sentence, " ")
 
 	// 使用 strings.Fields 按空格拆分成单词

static/index.html

@@ -229,6 +229,8 @@ <h1>世界帧精彩</h1>
 					<option value="zh_cn">简体中文</option>
 					<option value="en">英文</option>
 					<option value="ja">日文</option>
+					<option value="tr">土耳其语</option>
+					<option value="de">德语</option>
 				</select>
 			</div>