A Complete Guide to Ruby and Accent Marking in Narration Scripts: Tokyo vs. Keihan Patterns and Accurate Readings for IT, Medical, and Legal Terms

Ruby and Accent Instructions Are Not “Nice to Have” but Risk Control

Many mistakes in narration recording are not about performance but about inconsistent readings. In corporate videos, e-learning, medical explainers, and investor relations content, a wrong reading can immediately damage credibility. That is why ruby and accent instructions in the script matter.

In practice, it is not enough to add ruby only to difficult kanji. A reliable script separates instructions into at least these three categories:

1. Ruby to fix the reading of a term
2. Accent guidance to show the pitch pattern
3. Notes for industry-specific conventional readings

For example, the Japanese term for “optimization” may be straightforward in general usage, but in IT contexts the preferred pronunciation may follow internal convention or product branding. Most of these mismatches can be prevented before recording if the script is prepared properly.

Tokyo and Keihan Accent Systems Differ in Operational Standards, Not in “Correctness”

The first thing to understand is that Tokyo-style and Keihan-style accent systems are different frameworks, not a matter of right versus wrong. For nationwide distribution, broadcast work, and most corporate narration, Tokyo-style accent is the default. For Kansai-focused commercials, regional projects, or cases where local speech identity should remain, Keihan-style awareness becomes important.

A common mistake on the client side is giving vague directions like “make it sound Kansai-like” or “use natural standard Japanese.” That creates uncertainty in the booth. In real projects, write the rules explicitly:

• Standard: Tokyo-style
• Exception: Keihan-style allowed only for place names and proper nouns
• Supervision priority: client’s in-house pronunciation comes first
• If uncertain: prioritize internal audio references over dictionary entries

These four lines alone can reduce retakes significantly. In practice, cross-checking with the NHK Japanese Pronunciation Accent Dictionary, a major Japanese dictionary, and the client’s past videos works very well. I recommend narrowing questionable terms down to no more than 10 items before recording and confirming those first. Once the list exceeds 30 items, response speed from the client usually drops.

Industry-Specific Readings Should Follow Real Usage, Not Just Dictionaries

IT, medical, and legal fields are the three major areas where dictionary readings and real-world usage often diverge.

IT

In IT, loanwords are heavily localized, and usage varies by company. Terms such as “AWS,” “Azure,” “SQL,” and “GUI” may be read as letter names, abbreviations, or product-style pronunciations. The best method is to manage three columns:

• official name
• in-house common reading
• adopted reading for this project

Examples:

• SQL: “S-Q-L” / “sequel”
• GUI: letter-by-letter / conventional spoken form
• Azure: one of multiple accepted Japanese pronunciations depending on company style

Medical

In medical narration, the reading often changes depending on whether the audience is patients or healthcare professionals. A technical term may be read in a specialist way for clinicians but replaced with a plain-language reading for the public. Drug names and disease names are especially sensitive; one syllable off can point to something entirely different. These should always be checked with a supervisor.

Legal

In legal content, statutory reading, practical reading, and plain explanatory reading may all differ. Even professionals can disagree on accent patterns for legal terms. In legal narration, punctuation placement also affects comprehension. So besides accent, breath and pause positions should be marked as well.

A Practical Script Format You Can Use Immediately

The format I recommend is a two-layer structure that avoids overloading the main script.

Main script layer

• optimized for readability
• only essential ruby
• emphasis, pauses, and line breaks included

Annotation layer

• terminology list
• accent instructions
• exception rules
• reference audio links or internal glossaries

Example notation:

> 最適化（reading）[Tokyo-style, head-high]
> SQL（adopted reading for this project）
> 浮腫（public-facing reading） / specialist version uses another reading

For management, use Excel or Google Sheets with five columns:

term | ruby | accent | reason for adoption | supervisor approved

If the narration runs longer than about five minutes, keeping this table as a separate sheet noticeably improves recording efficiency.

The Five Minimum Items Every Client Should Provide

Finally, here are the five essentials that should be included when ordering narration:

1. Accent standard: Tokyo-style or Keihan-style
2. Priority order: dictionary, in-house convention, or supervisor instruction
3. Industry term list: provide up to 10 key terms in advance
4. Target audience: general public, professionals, or internal staff
5. Audio reference: past videos or sample URLs

Before a narrator can read beautifully, they must read correctly. Ruby and accent instructions are not fussy details; they are the blueprint that protects final quality. If the request document is organized even into a single solid page, retakes, recording time, and editing costs will all go down. In narration, reading accuracy is determined by script accuracy.