Skip to content

Formerly KrillinAI, VideoTranslation and Dubbing Tool for 120+ Languages. Open-source AI workspace for creators, powered by Codex. Video downloader, video translation, create videos, images, audios, and edits with Agents in one place. For your Social Media like YouTube, TikTok, Instagram, etc.

feat: single line subtitle & fix: failed to embed subtitle when use local file

沈浩189856ca60c7
diff --git a/internal/service/audio2subtitle.go b/internal/service/audio2subtitle.go
index 60cfe979..a2d4d7f0 100644
--- a/internal/service/audio2subtitle.go
+++ b/internal/service/audio2subtitle.go
@@ -3,6 +3,7 @@ package service
import (
"bufio"
"context"
+ "encoding/json"
"errors"
"fmt"
"krillin-ai/config"
@@ -216,11 +217,22 @@ func (s Service) audioToSrt(ctx context.Context, stepParam *types.SubtitleTaskSt
cancel()
return fmt.Errorf("audioToSubtitle audioToSrt splitTextAndTranslate err: %w", err)
}
+ _ = util.SaveToDisk(translatedResults, filepath.Join(stepParam.TaskBasePath, fmt.Sprintf(types.SubtitleTaskTranslationDataPersistenceFileNamePattern, translateItem.Id)))
log.GetLogger().Info("Translate completed", zap.Any("taskId", stepParam.TaskId), zap.Any("splitId", translateItem.Id))
- // 发送翻译结果
- translatedQueue <- DataWithId[[]TranslatedItem]{
- Data: translatedResults,
- Id: translateItem.Id,
+ // 分割长句
+ splitResults, err := s.splitTranslateItem(translatedResults)
+ if err != nil {
+ // 不中断
+ log.GetLogger().Error("audioToSubtitle audioToSrt splitTranslateItem err", zap.Any("taskId", stepParam.TaskId), zap.Any("splitId", translateItem.Id), zap.Error(err))
+ translatedQueue <- DataWithId[[]TranslatedItem]{
+ Data: translatedResults,
+ Id: translateItem.Id,
+ }
+ } else {
+ translatedQueue <- DataWithId[[]TranslatedItem]{
+ Data: splitResults,
+ Id: translateItem.Id,
+ }
}
}
}
@@ -1055,3 +1067,82 @@ func parseAndCheckContent(splitContent, originalText string) ([]TranslatedItem,
}
return result, nil
}
+
+// calcLength 计算文本视觉长度
+func calcLength(text string) float64 {
+ var length float64
+ for _, r := range text {
+ code := r
+ switch {
+ case (code >= 0x4E00 && code <= 0x9FFF) || (code >= 0x3040 && code <= 0x30FF): // 中日文
+ length += 1.75
+ case (code >= 0xAC00 && code <= 0xD7A3) || (code >= 0x1100 && code <= 0x11FF): // 韩文
+ length += 1.5
+ case code >= 0x0E00 && code <= 0x0E7F: // 泰文
+ length += 1
+ case code >= 0xFF01 && code <= 0xFF5E: // 全角符号
+ length += 1.75
+ default: // 其他字符(英文等)
+ length += 1
+ }
+ }
+ return length
+}
+
+// splitTranslateItem 根据字符权重和最大长度分割长句
+func (s Service) splitTranslateItem(items []TranslatedItem) ([]TranslatedItem, error) {
+ var result []TranslatedItem
+ maxLength := 70 // todo 先写死
+ //targetMultiplier := config.Conf.Subtitle.TargetMultiplier
+
+ for _, item := range items {
+ // 计算翻译文本的加权长度
+ if calcLength(item.OriginText) <= float64(maxLength) && calcLength(item.TranslatedText) <= float64(maxLength) {
+ result = append(result, item)
+ continue
+ }
+
+ // 调用大模型进行分割
+ log.GetLogger().Info("splitTranslateItem long sentence detected, need split", zap.Any("item", item))
+ splitItems, err := s.splitLongSentence(item)
+ if err != nil {
+ log.GetLogger().Error("splitTranslateItem splitLongSentence error", zap.Error(err), zap.Any("item", item))
+ return nil, fmt.Errorf("split long sentence error: %w", err)
+ }
+ result = append(result, splitItems...)
+ }
+
+ return result, nil
+}
+
+// splitLongSentence 使用大模型分割长句并保持原文和译文对齐
+func (s Service) splitLongSentence(item TranslatedItem) ([]TranslatedItem, error) {
+ prompt := fmt.Sprintf(types.SplitLongSentencePrompt, item.OriginText, item.TranslatedText)
+
+ response, err := s.ChatCompleter.ChatCompletion(prompt)
+ if err != nil {
+ return nil, fmt.Errorf("chat completion error: %w", err)
+ }
+
+ var splitResult struct {
+ Align []struct {
+ OriginPart string `json:"origin_part"`
+ TranslatedPart string `json:"translated_part"`
+ } `json:"align"`
+ }
+ if err := json.Unmarshal([]byte(util.CleanMarkdownCodeBlock(response)), &splitResult); err != nil {
+ log.GetLogger().Error("splitLongSentence parse split result error", zap.Error(err), zap.Any("response", response))
+ return nil, fmt.Errorf("parse split result error: %w", err)
+ }
+
+ // 转换为TranslatedItem切片
+ var splitItems []TranslatedItem
+ for _, part := range splitResult.Align {
+ splitItems = append(splitItems, TranslatedItem{
+ OriginText: part.OriginPart,
+ TranslatedText: part.TranslatedPart,
+ })
+ }
+
+ return splitItems, nil
+}
diff --git a/internal/service/link2file.go b/internal/service/link2file.go
index 40947588..f7f994e9 100644
--- a/internal/service/link2file.go
+++ b/internal/service/link2file.go
@@ -95,8 +95,8 @@ func (s Service) linkToFile(ctx context.Context, stepParam *types.SubtitleTaskSt
log.GetLogger().Error("linkToFile download video yt-dlp error", zap.Any("step param", stepParam), zap.String("output", string(output)), zap.Error(err))
return fmt.Errorf("linkToFile download video yt-dlp error: %w", err)
}
- stepParam.InputVideoPath = videoPath
}
+ stepParam.InputVideoPath = videoPath
// 更新字幕任务信息
stepParam.TaskPtr.ProcessPct = 10
diff --git a/internal/service/srt_embed.go b/internal/service/srt_embed.go
index 02ab0306..1c5ba6ef 100644
--- a/internal/service/srt_embed.go
+++ b/internal/service/srt_embed.go
@@ -229,20 +229,19 @@ func srtToAss(inputSRT, outputASS string, isHorizontal bool, stepParam *types.Su
if len(subtitleLines) < 2 {
continue
}
- var majorTextLanguage types.StandardLanguageCode
- if stepParam.SubtitleResultType == types.SubtitleResultTypeBilingualTranslationOnTop { // 一定是bilingual
- majorTextLanguage = stepParam.TargetLanguage
- } else {
- majorTextLanguage = stepParam.OriginLanguage
- }
+ //var majorTextLanguage types.StandardLanguageCode
+ //if stepParam.SubtitleResultType == types.SubtitleResultTypeBilingualTranslationOnTop { // 一定是bilingual
+ // majorTextLanguage = stepParam.TargetLanguage
+ //} else {
+ // majorTextLanguage = stepParam.OriginLanguage
+ //}
- majorLine := strings.Join(splitMajorTextInHorizontal(subtitleLines[0], majorTextLanguage, stepParam.MaxWordOneLine), " \\N")
- minorLine := util.CleanPunction(subtitleLines[1])
+ //majorLine := strings.Join(splitMajorTextInHorizontal(subtitleLines[0], majorTextLanguage, stepParam.MaxWordOneLine), " \\N")
// ASS条目
startFormatted := formatTimestamp(startTime)
endFormatted := formatTimestamp(endTime)
- combinedText := fmt.Sprintf("{\\an2}{\\rMajor}%s\\N{\\rMinor}%s", majorLine, minorLine)
+ combinedText := fmt.Sprintf("{\\an2}{\\rMajor}%s\\N{\\rMinor}%s", subtitleLines[0], util.CleanPunction(subtitleLines[1]))
_, _ = assFile.WriteString(fmt.Sprintf("Dialogue: 0,%s,%s,Major,,0,0,0,,%s\n", startFormatted, endFormatted, combinedText))
}
} else {
diff --git a/internal/types/embed_subtitle.go b/internal/types/embed_subtitle.go
index ddc7ac63..90293201 100644
--- a/internal/types/embed_subtitle.go
+++ b/internal/types/embed_subtitle.go
@@ -8,8 +8,8 @@ PlayDepth: 0
[V4+ Styles]
Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding
-Style: Major,Arial,18,&H00BFFF,&H000000FF,&H00000000,&H64000000,-1,0,0,0,100,100,0,0,1,2.5,1.5,2,10,10,20,1
-Style: Minor,Arial,12,&H00BFFF,&H000000FF,&H00000000,&H64000000,-1,0,0,0,100,100,0,0,1,2.5,1.5,2,10,10,30,1
+Style: Major,Arial,14,&H00BFFF,&H000000FF,&H00000000,&H64000000,-1,0,0,0,100,100,0,0,1,2.5,1.5,2,10,10,20,1
+Style: Minor,Arial,10,&H00BFFF,&H000000FF,&H00000000,&H64000000,-1,0,0,0,100,100,0,0,1,2.5,1.5,2,10,10,30,1
[Events]
diff --git a/internal/types/subtitle_task.go b/internal/types/subtitle_task.go
index 95c99b5d..33451220 100644
--- a/internal/types/subtitle_task.go
+++ b/internal/types/subtitle_task.go
@@ -118,6 +118,16 @@ var TranslateVideoTitleAndDescriptionPrompt = `你是一个专业的翻译专家
%s
`
+var SplitLongSentencePrompt = `请将以下原文和译文分割成2-3个部分,确保每个部分都尽可能短:
+原文:%s
+译文:%s
+
+要求:
+1. 分割后的句子必须保持语义完整,避免切断完整概念,避免切分后的句子不符合语法规范
+2. 确保原文和译文的分割部分一一对应
+3. 务必返回JSON格式,包含origin_part和translated_part数组,例如:
+{"align":[{"origin_part":"原文部分1","translated_part":"译文部分1"},{"origin_part":"原文部分2","translated_part":"译文部分2"}]}`
+
type SmallAudio struct {
AudioFile string
TranscriptionData *TranscriptionData
@@ -185,6 +195,7 @@ const (
SubtitleTaskTargetLanguageTextFileName = "target_language.txt"
SubtitleTaskStepParamGobPersistenceFileName = "step_param.gob"
SubtitleTaskAudioTranscriptionDataPersistenceFileNamePattern = "audio_transcription_data_%d.json"
+ SubtitleTaskTranslationDataPersistenceFileNamePattern = "translation_data_%d.json"
SubtitleTaskTransferredVerticalVideoFileName = "transferred_vertical_video.mp4"
SubtitleTaskHorizontalEmbedVideoFileName = "horizontal_embed.mp4"
SubtitleTaskVerticalEmbedVideoFileName = "vertical_embed.mp4"
diff --git a/pkg/util/base.go b/pkg/util/base.go
index c0635551..ff44b76e 100644
--- a/pkg/util/base.go
+++ b/pkg/util/base.go
@@ -191,6 +191,8 @@ func CopyFile(src, dst string) error {
// SanitizePathName 清理字符串,使其成为合法路径名
func SanitizePathName(name string) string {
+ name = strings.ReplaceAll(name, ".", "_")
+
var illegalChars *regexp.Regexp
if runtime.GOOS == "windows" {
// Windows 特殊字符
@@ -303,3 +305,10 @@ func LoadFromDisk(filename string) (any, error) {
err = decoder.Decode(&data)
return data, err
}
+
+// 清理 Markdown 的 ```json 标记
+func CleanMarkdownCodeBlock(response string) string {
+ re := regexp.MustCompile("(?m)^```(json|[a-zA-Z]*)?\n?|```$")
+ cleaned := re.ReplaceAllString(response, "")
+ return strings.TrimSpace(cleaned)
+}