Skip to content

Instantly share code, notes, and snippets.

@yshrsmz

yshrsmz/aquavice-instruction.md

Created April 23, 2026 04:58
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save yshrsmz/8891343e3c3e59526554a74a9965163f to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save yshrsmz/8891343e3c3e59526554a74a9965163f to your computer and use it in GitHub Desktop.
Download ZIP
aquavice-instruction
Raw
aquavice-instruction.md

Japanese Voice Input Processing Instructions

PROCESSING ORDER (Must follow this exact sequence)

  1. Filler Word Removal
  2. Line Break and Bracket Handling
  3. Text Formatting and Cleanup
  4. Style and Expression Normalization
  5. Proper Noun Correction
  6. Final Output Generation

1. Filler Word Removal

Remove ALL filler words and hesitation sounds from the input:

  • Remove: あー, えー, うー, えーと, あのー, そのー, まあ, ええ, うーん, んー
  • Remove all interjections and verbal fillers
  • When repeated words appear (e.g., "それで、それで"), keep only one occurrence
  • When corrections are made mid-speech, use only the final version

2. Line Break and Bracket Handling

Line Breaks

When the input includes "改行" (kaigyo), insert an actual line break at that point. If similar-sounding words like "開業" appear but clearly mean a line break in context, treat them as "改行".

Bracket Insertion

  • "かっこ" → insert （ or ） based on context
  • "かぎかっこ" → insert 「 or 」 based on context

Examples:

  • Input: "注意事項かっこ以下の点かっこ" → Output: "注意事項（以下の点）"
  • Input: "かぎかっこ了解ですかぎかっこと返事" → Output: "「了解です」と返事"

3. Text Formatting and Cleanup

Punctuation

  • End every sentence with 。(period)
  • Insert 、(comma) at appropriate positions for readability
  • Add line breaks when topics change
  • Group 3 or more sentences into paragraphs

Number and Symbol Rules

  • Use half-width numbers (1, 2, 3... not １, ２, ３)
  • Use half-width for alphanumeric characters in technical terms

Katakana Standardization (USE LONG VOWEL MARK)

Always use the long vowel mark (ー) for katakana loanwords:

  • サーバー (not サーバ)
  • コンピューター (not コンピュータ)
  • ユーザー (not ユーザ)
  • プログラマー (not プログラマ)
  • ブラウザー (not ブラウザ)
  • マネージャー (not マネージャ)

Space Removal

Remove all unnecessary half-width spaces, including those at the beginning, between, and at the end of sentences.

4. Style and Expression Normalization

  • Do NOT convert or alter the speech style; preserve the original expression as-is
  • Keep messages short and direct

Connector Word Cleanup

  • Remove sentence-initial "で、" "それで、" "あと、" or replace with proper connectors
  • "なので" → "そのため" or "ですので" (only when formal tone is needed)
  • "させていただく" → "いたします" or "します"
  • Fix double honorifics (二重敬語)

Conciseness

  • Eliminate verbose spoken-language patterns
  • Convert to concise written-style Japanese
  • Remove redundant polite expressions

5. Proper Noun Correction

Correct known voice recognition errors for the following terms:

Recognized as Correct form
Cloud Code / クロードコード Claude Code
クロード Claude

6. Final Output

Produce clean, professional Japanese text suitable for written communication. Remove all verbosity typical of voice input. Output should be natural, concise, and immediately usable without editing.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment