Little tool to fix overlapping subtitles (especially the ones extracted from english auto-subtitles of youtube, vtt files that you would convert to srt with ffmpeg)

subtitle-overlap-fixer.go

package main

import (
	"time"
	"regexp"
	"bufio"
	"strconv"
	"fmt"
	"os"
	"errors"
	"io"
	"strings"
)

type Subtitle struct {
	idx int
	fromTime time.Duration
	toTime time.Duration
	text string
}

var timeFramePattern, _ = regexp.Compile(`(\d+):(\d+):(\d+),(\d+) --> (\d+):(\d+):(\d+),(\d+)`)

func getDuration(parts []string) time.Duration {
	hour, _ := strconv.Atoi(parts[0])
	minute, _ := strconv.Atoi(parts[1])
	second, _ := strconv.Atoi(parts[2])
	millisecond, _ := strconv.Atoi(parts[3])
	return time.Millisecond * time.Duration(millisecond) +
		time.Second * time.Duration(second) +
		time.Minute * time.Duration(minute) +
		time.Hour * time.Duration(hour)
}

func printDuration(duration time.Duration) string {
	hour := duration / time.Hour
	duration -= hour * time.Hour
	minute := duration / time.Minute
	duration -= minute * time.Minute
	second := duration / time.Second
	duration -= second * time.Second
	millisecond := duration / time.Millisecond
	return fmt.Sprintf(`%02d:%02d:%02d,%03d`, hour, minute, second, millisecond)
}

func readOneSubtitle(scanner *bufio.Scanner) (*Subtitle, error) {
	// read idx
	if !scanner.Scan() {
		return nil, nil
	}
	idxRaw := scanner.Text()
	idx, err := strconv.Atoi(idxRaw)
	if err != nil {
		return nil, errors.New("invalid subtitle index")
	}
	// read timing
	if !scanner.Scan() {
		return nil, errors.New("could not find subtitle timing")
	}
	timing := timeFramePattern.FindStringSubmatch(scanner.Text())
	if timing == nil {
		return nil, errors.New("invalid subtitle timing")
	}
	fromTime := getDuration(timing[1:5])
	toTime := getDuration(timing[5:9])
	// read content
	if !scanner.Scan() {
		return nil, errors.New("could not find subtitle text")
	}
	content := scanner.Text()
	for scanner.Scan() && scanner.Text() != "" {
		content += "\n"
		content += scanner.Text()
	}
	subtitle := &Subtitle{idx, fromTime, toTime, content}
	return subtitle, nil
}

func writeOneSubtitle(file io.Writer, subtitle *Subtitle, idx *int) error {
	_, err := fmt.Fprint(file,
		*idx, "\n",
		printDuration(subtitle.fromTime), " --> ", printDuration(subtitle.toTime), "\n",
		subtitle.text, "\n\n")
	*idx++
	return err
}

func main() {
	if len(os.Args) < 2 {
		println("Provide a subtitle file to fix.\ne.g. subtitle-fixer mysubtitle.srt")
		return
	}

	filePath := os.Args[1]
	newFilePath := filePath + ".fixed"

	file, _ := os.Open(filePath)
	newFile, _ := os.Create(newFilePath)
	defer file.Close()
	defer newFile.Close()

	scanner := bufio.NewScanner(file)
	var newIdx = 1
	var lastSubtitle *Subtitle = nil

	for {
		subtitle, err := readOneSubtitle(scanner)
		if lastSubtitle != nil {
			if subtitle != nil {
				subtitle.text = strings.Trim(subtitle.text, "\n ")
				if len(subtitle.text) == 0 { // skip over empty subtitles
					continue
				}
				// skip over super-short subtitles that basically contain what their previous subtitle contains, and just prolong previous subtitle
				if subtitle.toTime - subtitle.fromTime < time.Millisecond * 150 &&
					strings.Contains(lastSubtitle.text, subtitle.text) {
					lastSubtitle.toTime = subtitle.toTime
					continue
				}
				// if first-line of current subtitle is repeating last-line of previous-subtitle remove it
				currentLines := strings.Split(subtitle.text, "\n")
				lastLines := strings.Split(lastSubtitle.text, "\n")
				if currentLines[0] == lastLines[len(lastLines)-1] {
					subtitle.text = strings.Join(currentLines[1:], "\n")
				}
				// if first-line of current subtitle is repeating last-line of previous-subtitle remove it
				if subtitle.fromTime < lastSubtitle.toTime {
					lastSubtitle.toTime = subtitle.fromTime - time.Millisecond
				}
			}
			writeOneSubtitle(newFile, lastSubtitle, &newIdx)
		}
		if subtitle == nil {
			break
		}
		if err != nil {
			panic(err)
		}
		lastSubtitle = subtitle
	}

	os.Rename(filePath, filePath + ".bak")
	os.Rename(newFilePath, filePath)
}

@niceindividual That would be my pleasure. 🌹

For anyone else who would find this useful:

I had an issue with my SRT file being rejected by a picky program for having occasional blank entries like

42
00:03:14,000 --> 00:03:14,159

I moved the section at https://gist.github.com/nimatrueway/4589700f49c691e5413c5b2df4d02f4f#file-subtitle-overlap-fixer-go-L111-L113 down to the end of the block to perform this check last. My thinking is that I was running into issues at line https://gist.github.com/nimatrueway/4589700f49c691e5413c5b2df4d02f4f#file-subtitle-overlap-fixer-go-L124 that subverted the previous empty line check. In any case, pushing this section down seemed to resolve my issues!

Cheers to @nimatrueway for this awesome script. It saved me a ton of time!

Great job! It solved the problem I had with the overlapping. Thanks a lot.

AWESOME! Did exactly what I needed. There were some additional fixes I needed to do (sentence capitalization and changing lowercase "i" to uppercase "I" where needed.). I wrote a little OS X bash script for these issues if anyone want to try it. https://github.com/bruno-sardine/mac#Further-correct-YouTube-captions-captfixsh

For anyone else who would find this useful:

I had an issue with my SRT file being rejected by a picky program for having occasional blank entries like

42
00:03:14,000 --> 00:03:14,159

I moved the section at https://gist.github.com/nimatrueway/4589700f49c691e5413c5b2df4d02f4f#file-subtitle-overlap-fixer-go-L111-L113 down to the end of the block to perform this check last. My thinking is that I was running into issues at line https://gist.github.com/nimatrueway/4589700f49c691e5413c5b2df4d02f4f#file-subtitle-overlap-fixer-go-L124 that subverted the previous empty line check. In any case, pushing this section down seemed to resolve my issues!

Cheers to @nimatrueway for this awesome script. It saved me a ton of time!

Thanks fork it

Thank you for sharing this, @nimatrueway.

Sir, I compiled your original code and tested it by fixing a subtitle file, say abc.srt. The output ends up being named abc.srt.fixed, which is a bit inconvenient since the operating system doesn't recognize it as a standard subtitle file due to the altered extension. So I modified the code to generate the output as abc.fixed.srt instead—this way, the .srt extension remains intact and the file is still recognized properly.