THE SHORT ANSWER

Check the source language first: Japanese audio transcription is not supported. For supported material, agree on a separate subtitle file or captions displayed in the video, correct the words, then test a short export in the actual destination player.

BEFORE YOU READ

Who this research is for

Read now

  • You need captions or subtitle files for supported-language media
  • You want an explicit correction and delivery review process

Check something else first

  • Automatic transcription of Japanese recordings is the main requirement
  • Nobody can review the source wording or translation

KEY POINTS

What matters most

  1. 01

    Caption translation into Japanese does not mean Japanese audio transcription

  2. 02

    Choose subtitle files or on-video captions from the delivery specification

  3. 03

    Review meaning, timing and the final export instead of only the editor preview

DECISION TABLE

Choose the deliverable before styling captions

Choose the deliverable before styling captions
Recipient requirementOutput to evaluateAcceptance check
Separate video and subtitle assetsSRT or VTTThe specified format matches the final video version
A consistent visible caption treatmentVideo with a captions layerText is readable in the destination player
Another audience languageReviewed translated captionsNames, numbers, conditions and negatives retain their meaning
Japanese source audioA different transcription workflowConfirm source-language support before uploading

Follow the recipient's specification. A subtitle file is not a video file; test its appearance and timing where it will actually be used.

ACTION PLAN

Test a short deliverable in four stages

These are editorial acceptance criteria, not promises of automatic accuracy. Finish one small sample in the real destination before scaling up.

  1. 01

    Write the delivery specification

    Record the source and caption languages, file format, aspect ratio, resolution and receiving platform. Resolve ambiguous requirements with the recipient before production.

  2. 02

    Edit and correct a permitted sample

    Use one or two minutes of representative material. Finalize the speech selection, then check names, numbers and negatives against the audio. Replay joins after edits.

  3. 03

    Export one version at a time

    Approve the original-language output before translation. Give the video and subtitle files matching revision identifiers and check for duplicated caption display.

  4. 04

    Record the destination test

    Check the beginning, edited joins and ending for synchronization, clipped text, line breaks, audio and watermark. Log failures by timecode and replay the corrected segment with its surrounding context.

WORKED EXAMPLE

Worked example: a three-minute video needs its own review budget

The following numbers are assumptions for planning, not Descript performance measurements or plan limits.

Material
Edit 12 minutes of source into a 3-minute video
Assumed first review
8 minutes checking words + 4 checking caption display + 3 testing delivery = 15 minutes
Second subtitle language
Budget a separate language review; output duration alone is not the workload
Later source edit
Recheck the affected captions, translation and exported file pair

What this example showsMeasure your own time by stage. Before increasing output volume, assign responsibility for language review, corrections and replacing delivered files.

01

Separate source-language support from translated output

Descript's documentation excludes Japanese audio from automatic transcription while listing Japanese as a caption-translation destination. Treat these as different stages. This workflow is aimed at supported source material such as English, not at automatically transcribing a Japanese meeting.

Write down the language spoken in the recording before selecting software. If it fails that entry condition, change the workflow before estimating editing time. A translated output option is not evidence that the same language can be used as input.

02

Correct words without accidentally cutting speech

Use Correct to repair transcription errors while keeping the underlying recording. Ordinary script deletion is a media edit. Decide which speech belongs in the finished piece before doing a detailed pass through the caption wording.

Our suggested review order prioritizes names, amounts, dates, units and negative statements. If the speaker made a factual mistake, silently replacing the caption with a different claim creates an audio-text mismatch. Decide whether to add an explicit correction or explanatory note instead.

03

Keep caption layers and their scope under control

Use the Captions panel to apply a style to the intended scene range. Select the existing captions layer when changing its style so that you do not unintentionally add another layer.

For acceptance, inspect long sentences, speaker changes and scenes containing charts. Look for text covering faces, figures or the destination player's controls. Judge readability on the actual device as well as in the editor; a desktop preview is not the recipient's viewing environment.

04

Deliver subtitle files with the matching video revision

The subtitle exporter supports SRT and VTT, with controls for speaker labels and line layout. Use the format the recipient requests. Styling a captions layer does not by itself deliver a separate subtitle file.

We suggest matching versioned names such as clip-v03.mp4 and clip-v03.en.srt. If you cut the video after generating the subtitles, check the pair again. Tell the recipient which files replace the old version instead of relying on filenames such as final-final-new.

05

Review translations after source changes

Translated captions live in separate compositions. Changes to the original do not synchronize automatically, and retranslation can overwrite manual corrections. Stabilize the original before translating and record later changes.

Review who did what, dates, quantities, negation and conditional claims against the source. Fluent wording is not sufficient evidence of accuracy. Include a qualified language review in the schedule when you cannot evaluate the translation yourself.

06

Run one short export before committing the full job

Use our plan comparison for subscription boundaries, then test the required subtitle format, video resolution, watermark and translation entitlement in your own account. Keep these as separate acceptance checks. A successful import is only the beginning of the workflow.

When a sample fails, classify the cause: plan restriction, mistaken operation, unsupported source language or unclear source material. Upgrade only when the relevant plan difference addresses that cause. Do not treat every failure as a reason to buy a more expensive subscription.

COMMON PITFALLS

Where decisions go wrong

01

Accepting appearance without checking meaning

One missing negative, extra zero or wrong unit can invalidate an otherwise polished caption. Treat consequential errors as failures.

02

Combining different video and subtitle revisions

Before manually adjusting timestamps, verify that both files represent the same edit. Otherwise you may repair the wrong pair.

03

Losing corrections during retranslation

Record source changes and manual translation fixes before refreshing the translation, then inspect the result instead of assuming corrections survived.

DECISION NOTE

Keep a video-to-subtitle handoff record

Record the video file, subtitle file, source language, caption language, revision, reviewer and unresolved issues together. Acceptance means the recipient can use that exact version, not merely that an export completed. This record also identifies which outputs require another translation or export after a source edit.

CLAIM → SOURCE

Claims and supporting sources

Each central claim points to a primary or public source reviewed for this article.

  • Japanese audio transcription and Japanese caption translation have different support

  • Correct mode differs from deleting media through script edits

  • Caption layers and separate SRT/VTT exports are different delivery options

  • Source changes do not automatically synchronize to translations

FAQ

Frequently asked questions

Q1

Can I automatically transcribe Japanese audio in Descript?

That input is outside this documented workflow. Do not infer input support from a Japanese caption-translation option; choose a supported transcription workflow first.

Q2

Do I need both SRT and captions inside the video?

Only if the recipient requests both. Agree whether captions should be controlled by the player or remain visible in the video, and test for duplicated display.

Q3

Will fixing a caption remove the spoken word?

Use Correct for a text repair and keep it distinct from a media cut. Replay the edited segment to confirm the intended speech remains.

Q4

Can Free finish my deliverable?

That depends on the required output format, video resolution, watermark and translation features. Export a small sample in your account and compare it with the recipient's specification before proceeding.

METHODOLOGY

How this article was researched

We checked six Descript help pages on September 10, 2026. The review sequence, acceptance criteria and worked example below are editorial proposals, not hands-on accuracy or speed measurements.

Read the shared editorial method ↗

SOURCES

Sources reviewed