A Two-Pass Subtitle Workflow: Draft, QA, and Polish

Begin with an existing timed SRT and the matching video cut. In the first pass, produce a source-accurate dialogue draft. In the second, review that draft against the scene, revise the language, and inspect the exported subtitles during playback.

The useful separation is between establishing what the dialogue means and deciding how it should read on screen. Each pass ends with a specific handoff, so unresolved meaning does not disappear inside a polished sentence.

Set up the working file

Keep an unchanged copy of the supplied SRT. Confirm its video version and note any missing exchanges or source-file discrepancies before translating.

Have the approved names and terms, character voice notes, and delivery requirements beside the working file. Keep a separate query log with the cue reference, speaker, uncertainty, and decision needed. This log carries editorial questions between passes; it does not belong in the delivered subtitle text.

beside the working file

a separate query log

the approved names and terms, character voice notes, and delivery requirements

the cue reference, speaker, uncertainty, and decision needed

This log carries editorial questions between passes

it does not belong in the delivered subtitle text.

If you are also responsible for creating the original timing from audio, complete that preparation in an external subtitle editor before beginning this workflow. The complete short-drama guide covers the wider source-to-delivery process.

Pass 1: Build a source-accurate draft

Work through each exchange with its surrounding scene. Confirm the speaker, listener, action, and information available to the audience. Preserve who did what, negation, conditions, relationships, and plot references.

Translate the function as well as the facts. A warning, reluctant agreement, and sarcastic compliment can contain similar words but perform different actions. Use the source audio, performance, and next reply to resolve the intended reading.

Use established names and terminology. When a new term appears, record a provisional choice and its context rather than silently introducing an alternative. Mark uncertainty about an idiom, reference, or relationship in the query log.

Draft into the existing cue structure. Keep this pass focused on the dialogue; record suspected timing or segmentation problems for manual review in your external subtitle editor during the second pass. Flag long drafts for adaptation while retaining their plot facts.

Preserve

Look for

who did what, negation, conditions, relationships, and plot references

omitted dialogue, repeated text, missing negation, and accidental changes in who is speaking or being addressed

The first-pass handoff

• the draft SRT
• the query log

Before handing over, compare the draft with the source cue by cue. Look for omitted dialogue, repeated text, missing negation, and accidental changes in who is speaking or being addressed.

The first-pass handoff is the draft SRT plus the query log. It should make clear which lines have settled meanings and which need a source decision. If you are working alone, keep that distinction when you return for review.

Pass 2: Review, adapt, and test in context

Resolve meaning and polish the language

Resolve source questions before committing to a shorter line. Then read each exchange aloud and compare it with the performance. Replace stiff sentence structure, remove unnecessary repetition, and preserve the words carrying the action or emotion.

The following invented English dialogue is a hypothetical text-only editorial illustration, not a source-language translation or product output. In this scene, an angry speaker threatens a rival and the rival’s company:

The revision keeps the condition and both targets of the threat. Cutting it to “Stop now, or you’ll pay” would lose the reference to the company. Check every shortening against the source with the same attention to what remains.

Review character register and recurring terms next. A shift from formal to familiar speech may be motivated by the scene; check it against the voice notes and relationship context. Compare callbacks with their earlier uses before standardizing them.

Review cue structure manually in a subtitle editor

Load the working SRT with the matching video in your external subtitle editor. Inspect cue entrances and exits, speaker changes, reveal order, and cuts. If you retime a cue, check the surrounding exchange again rather than judging the edited cue in isolation.

In that editor, compare line length and reading speed with the delivery platform’s documented requirements. Characters per line measure a line’s text length; characters per second are the cue’s character count divided by the number of seconds it is displayed. Netflix’s U.S. English subtitle guide uses these measures for translated subtitles delivered to Netflix in U.S. English; its reading-speed FAQ includes spaces and punctuation in the count. Use your destination’s target-language and audience policy, then review wording complexity and the action on screen.

line length and reading speed

Characters per line

a line’s text length

characters per second

the cue’s character count divided by the number of seconds it is displayed

Use your destination’s target-language and audience policy

In the external editor, try breaks at grammatical boundaries. For English, keeping nouns with their articles or adjectives and main verbs with their auxiliaries follows Netflix’s General Requirements, §4, used for timed text delivered to Netflix together with the language-specific guide. Keep other phrases together where the meaning and delivery layout allow it. The vertical-video line-break guide gives examples of these decisions.

When shortening cannot preserve a necessary idea, review cue splitting or merging manually in the external editor. Check that the new structure still follows the dialogue and does not expose information before its source moment.

Watch the exported subtitles in the delivery player

Export the revised SRT from the external editor and watch it with the video in the actual delivery player. For vertical video, include the intended phone display. Inspect font size, wrapping, contrast, subtitle position, and collisions with controls or on-screen text.

Watch continuously before stopping for corrections. Mark moments where reading competes with the performance, a cue arrives late, or a line sounds wrong for the exchange. Return to the source, make the correction in the working file, export again, and replay the affected sequence.

Watch continuously before stopping for corrections.

1. Return to the source
2. make the correction in the working file
3. export again

replay the affected sequence.

Close the review with a clear handoff

The second pass is complete when the source questions are resolved, the edited wording has been compared with the scene, and the final export has been reviewed in the delivery player.

Keep the delivered SRT with its matching video version, updated term decisions, and review record. Check that editorial comments have stayed in the query log and that the delivered file is the export you actually watched. If a later source change affects a cue, reopen the relevant meaning and playback checks for that exchange.

The second pass is complete when

the source questions are resolved

the edited wording has been compared with the scene

the final export has been reviewed in the delivery player

Keep the delivered SRT with

• its matching video version
• updated term decisions
• review record