fix(lyrics): preserve Telugu shaping in karaoke

[Alvaro-Manzo]

Fixes #1084.

Timed TTML can split Telugu words into multiple karaoke fragments. Rendering those fragments separately breaks Telugu text shaping and can cause overlapping glyphs.

This change groups fragments belonging to the same Telugu word before creating karaoke syllables, while preserving word-level timing and phonetics.

[Copilot]

Pull request overview

This PR improves Enhanced Lyrics TTML karaoke rendering for Telugu by grouping TTML fragments into larger layout nodes so Telugu grapheme shaping is preserved and glyph overlap is avoided.

Changes:

• Group WordTimestamp fragments for Telugu into “syllable groups” before building KaraokeSyllables.
• Merge grouped fragment text into a single KaraokeSyllable.content while keeping timing derived from the group boundaries.
• Aggregate per-fragment romanization (“phonetics”) into a single phonetic string per grouped syllable.

1. PR Summary

The grouping approach is directionally correct for fixing Telugu shaping issues caused by TTML fragmenting. However, the current phonetic aggregation inserts spaces between fragments, which can break word-level romanization, and the phonetic index computation is O(n²) in the common (non-Telugu) case.

2. Blocking Issues (Changes Requested)

• Grouped fragment phonetics are joined with spaces, which can introduce incorrect word breaks in romanization and conflicts with the PR’s intent to preserve word-level phonetics.

3. Non-Blocking Feedback (Nitpicks & Polish)

• phoneticStartIndex is recomputed via prefix summing inside mapIndexed, making toKaraokeSyllables() O(n²) when no grouping occurs (typical for non-Telugu). This is avoidable with a small refactor.

4. Suggested Code Fixes

app/src/main/kotlin/moe/rukamori/archivetune/ui/component/LyricsEnhanced.kt

• Concatenate grouped fragment phonetics without separators:

.filterNotNull()
.joinToString(separator = "")
.takeIf(String::isNotEmpty)

• Precompute phonetic offsets once instead of summing inside mapIndexed:

val phoneticOffsets =
    IntArray(syllableGroups.size).also { offsets ->
        var offset = 0
        syllableGroups.forEachIndexed { i, group ->
            offsets[i] = offset
            offset += group.size
        }
    }

// ...
val phoneticStartIndex = phoneticOffsets[index]

5. Final Verdict

REQUEST CHANGES

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.