Related issue

• Fixes:Subtitles appear twice sometimes NewPipe#11204
• Subtitles not shown properly #1102

Scope of changes

This PR involvestwo repositories:

• NewPipeExtractor (main changes)
  Implements subtitle deduplication logic in
  SubtitleDeduplicator.

• NewPipe (supporting changes)
  Initializescache/subtitle_cache directory and ensures locally cached
  subtitle files can still be manually downloaded as '*.srt'.

Reproduction case

Android device, duplicated subtitles visible during playback

YouTube video used for testing:
https://www.youtube.com/watch?v=b7vmW_5HSpE

Subtitle cache location

Cached subtitle files (*.ttml) are stored at:

/storage/emulated/0/Android/data/<package_name>/cache/subtitle_cache

The directory name corresponds tosubCacheDir
defined inSubtitleDeduplicator.

Cache file naming

Cached subtitle filenames are intentionally descriptive,
so their meaning can be understood without reading the code
(e.g. source, format, origin, subtitle state). For example:
cache/subtitle_cache $ ls -l
total 48
-rw-rw---- 1 u0_a579 sdcard_rw 1214 2026-01-29 17:44 b7vmW_5HSpE--en--auto_generated--original.ttml
-rw-rw---- 1 u0_a579 sdcard_rw 42426 2026-01-29 17:44 b7vmW_5HSpE--en-GB--human_provided--deduplicated.ttml

Cache lifecycle & storage impact

Do cached subtitle files need to be deleted?

No.

SubtitleDeduplicator does not delete cached subtitle files,
regardless of whether duplication is detected.

Why keep cached subtitles?

1. If a remote subtitle download fails, a previously cached version
   can be reused.

2. In practice, download failures are rare:

   • User-uploaded and auto-generated YouTube subtitles were
     consistently downloadable in tests.
   • Auto-translated subtitles showed a higher failure rate,
     but that feature is not merged intodev branch and is out of scope here.

Storage considerations

• Subtitle files are small in size.
• Even with many cached subtitles, storage usage grows slowly.
• In the worst case, Android will notify users of low storage
  and suggest clearing app cache, which includes NewPipe’s subtitle cache.

Unit tests

extractor/src/test/java/org/schabi/newpipe/extractor/utils/SubtitleDeduplicatorTest.java

Tests focus on thecore deduplication logic:
detecting duplicated adjacent subtitle segments and verifying
the resulting output.

Why SubtitleDeduplicator operates on raw TTML text

SubtitleDeduplicator intentionally operates on raw TTML textbefore XML entity decoding.
Deduplication is limited to lightweight, string-level normalization to avoid double
subtitle parsing into the screen-display and SRT manually download layer.

This design is intended to be practical and simple. At this stage, the goal is only to detect obviously
duplicated subtitle segments from the same TTML source, not to fully interpret
or normalize subtitle semantics.

Difference fromSrtFromTtmlWriter

These two components serve different purposes:

• SrtFromTtmlWriter

  • Performs full TTML XML parsing
  • Decodes entities
  • Resolves tags (<span>,<br>, etc.)
  • Generates SRT for manual download

• SubtitleDeduplicator

  • A lightweight pre-processing utility
  • Doesnot parse XML
  • Doesnot decode entities
  • Performs minimal string-level normalization only

Note on subtitle caching

SubtitleDeduplicator always fetches remote subtitle content to ensure the latest version is used when detecting duplicated entries.

During playback, however, ExoPlayer may serve subtitle data from its internal cache (cache/exoplayer) if a cached version is available. As a result, there is a potential inconsistency where the subtitle content displayed to the user may not immediately reflect a recently updated remote subtitle.

This is intentional and won’t be changing for now:

• YouTube subtitles don’t update often, so while users might encounter outdated cached subtitles, the chance is really low.
• ExoPlayer’s caching is part of the player, and I’m not sure how much code changing it would require, but that’s outside the scope of this PR.
• This PR is all about fixing subtitle duplication, not touching the caching setup between the extractor and player.

The fix has been tested with a YouTube video link:https://www.youtube.com/watch?v=b7vmW_5HSpE

Before the fix, the subtitle is shown as follows:

After applying the fix, the subtitle is displayed as follows:

@AudricV

Thanks for the feedback, it’s helpful for me to better understand the intended boundaries of NewPipeExtractor.

I’m preparing some follow-up comments to explain these commits, especially around subtitle downloading. I’m also taking some time to think about whether this design makes sense.

I’ll add more comments soon.

About

“we don't want NewPipe Extractor to download files directly”

:

@AudricV

Just to make sure I understand correctly: currently, the extractor only provides
subtitle URLs, and the actual downloading is done later on the app side
(either by ExoPlayer or by the manual subtitle download feature), right?

My original idea was to fix duplicated subtitles as early as possible and
in a centralized place — at the source where subtitle URLs are produced.
That’s why I chosegetSubtitles(final MediaFormat format) in
YoutubeStreamExtractor.java. If the source is deduplicated, all later code
would receive clean subtitles.

However, I now realize that my changes effectively moved the subtitle downloading
responsibility. Previously, subtitle URLs were passed through the extractor and
only downloaded on the app side. With this change, subtitles are downloaded
insideSubtitleDeduplicator instead.

At first, I thought this was acceptable since subtitles are eventually downloaded
anyway, and this could even reduce network requests by avoiding separate downloads
for playback (via ExoPlayer) and for manual SRT downloads. But after tracing the
code path more, I see that fromgetSubtitles() in NewPipeExtractor to
VideoPlaybackResolver.resolve() on the app side, subtitles are still handled
purely as URLs, without any download happening in the extractor.

So, performing file downloads inside NewPipeExtractor crosses its intended boundary, right?

Quality Gate passed

Issues
20 New issues
0 Accepted issues

Measures
3 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud