PasteLint
← Back to Text Preparation Journal

Engine Room

Preserving Catalog Records Across SSML Chunks

By Guy Teichman

Long audio scripts are not just text. They are production sequences. If a book title starts in one chunk and its details continue in the next, the script can sound broken even when no words were technically lost.

Chunk boundary

Title DB number Duration Author Narrator
record-aware chunking speech review audio prep

A chunk boundary is safer when it respects the catalog record instead of slicing through the middle of it.

The problem

A title at the end of one chunk can orphan the rest of the record.

SSML Builder can split long scripts into smaller chunks for audio production. That is useful, but chunk boundaries matter. A chunk should not casually begin the next book record at the very end if the title's DB number, author, narrator, description, or source note continues in the next chunk.

Compact example

Before: Chunk 2 ends with: Jump and Find Joy: Embracing Change in Every Season of Life, DB 134076...

Chunk 3 continues with: read by Hoda Kotb. Description...

After: The full catalog record starts together in the next chunk when it fits.

Why it matters

For speech and IVR, chunks are production boundaries.

A bad boundary can make a script harder to review, harder to record, and easier to misunderstand. The words may still be present, but the person reviewing the audio script has to reconnect pieces that should have stayed together.

What changed

SSML Builder now treats catalog records as stronger units.

  • SSML Builder looks for catalog-record patterns.
  • DB numbers, duration, author, and narrator metadata are treated as strong clues.
  • When a full record fits under the chunk limit, it stays together.
  • Oversized records still split only as a fallback.
  • The existing 3000-character safety limit remains unchanged.

Regression coverage

The fix is checked against the record, not only the character count.

  • The title appears in full.
  • Metadata stays with the record when it fits.
  • Chunks stay within the limit.
  • Text is not dropped, duplicated, or truncated.

Text-prep principle

A chunk boundary is an editorial decision.

Splitting text is not just a technical action. In speech workflows, the split changes how a human reviews and records the script.

Related PasteLint tools

Useful when long scripts need to stay reviewable.

Speech markup

SSML Builder

Prepare reviewed scripts for speech markup and chunked audio production.

Phone scripts

IVR Text Prep

Clean phone menu text before it moves into speech or phone-system review.

Read-aloud prep

TTS Text Cleanup

Prepare text so read-aloud output is easier to review before SSML handoff.

Source signal

Seen a cleanup edge case?

Send a source signal or bug example so future Engine Room notes can document what changed and why.

Share a source signal

Related reading

Structure can carry meaning.