docs: preserve jokes/puns + homophone-flatten footgun - #5
Merged
Conversation
Comedy skits live or die on the pun, and Whisper silently normalizes a homophone pun to the common word — so the gag never reaches the transcript and no proper-noun fix recovers it. Add: - step 3: preserve the joke (trans-create puns; the subtitle is the joke's primary carrier, since the dub rides over the ducked original). - a footgun row: the burned-in caption spells out the pun character, so OCR is usually the only reliable recovery path. Learned localizing a real Mandarin skit: 人工智能 (AI) -> 人工痔能 and 搁置 (shelve) -> 割痔疮 (hemorrhoid removal), both flattened by ASR, both recovered from the on-screen caption. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
TL;DR
Comedy/skit inputs live or die on the pun, and the skill said nothing about preserving jokes. Worse, homophone puns are invisible to the ASR — Whisper "corrects" the rare pun to the common word, so the gag never reaches the transcript. This adds a step-3 clause + one footgun row. Diff is +2/−1 lines.
Verdict: learned from a real localization this session; the pun-flattening was confirmed against the video's own burned-in caption (not inferred). Chinese + non-redundancy checked by a fresh lineage (Codex).
What changed
The real case that earned this
A Mandarin office-revenge skit (
youtube.com/shorts/wh2_aBn_3MQ). Two homophone puns bookend it, and Whisper flattened both:智and痔are bothzhì;搁置≈割痔exactly in sound, with疮appended. Neither pun surfaced in ASR; both were recovered by OCR'ing the caption band. The English localization rendered them as "Artificial Intelligence" → "Artificial PAIN-telligence" and the finale as "sit on it" (shelve) → hemorrhoids so the joke survives with zero Chinese knowledge.Fresh-lineage review (Codex, gpt-5.x)
智/痔bothzhì; 人工智能=AI; 人工痔能=valid character-substitution pun; 搁置=shelve; 割痔疮=hemorrhoid removal).Disjoint from open PR #4; branches off
main.