The contrarian take: AI doesn't replace your localization team
AI subtitle translation should not replace translators and QC editors. For agency work, that is the wrong goal.
The better use is cutting down the first-draft grind: the plain pass where every line gets translated before the real craft begins.
Your team still owns tone, terminology, client rules, QC, and final sign-off. Mitsuko gives them a context-aware draft to start from, not a black-box final delivery.
What "context-aware" means for an agency file
The difference between a generic machine draft and a usable agency draft is whether the model knew the project rules before it started.
Here is the same Korean drama line handled both ways.
Project terminology (user glossary):
도련님= Mr. Park,한강그룹= Han River Holdings. The speaker is a secretary addressing the heir of a family-owned company.
| Type | Line |
|---|---|
| Korean drama sample line | 도련님, 한강그룹 사람들이 벌써 로비에 와 있습니다. |
| Literal machine translation | Young master, Hangang Group people already to the lobby have come. |
| Context-aware translation (Mitsuko) | Mr. Park, the Han River Holdings representatives are already waiting in the lobby. |
The literal version breaks three things at once. It translates 도련님 to "Young master" instead of the client's glossary entry "Mr. Park."
It also changes 한강그룹 into "Hangang Group" and produces English no human would write.
The context-aware version follows the glossary, uses the role context (secretary to heir), and reads like a line an editor can work with instead of rewrite.
Series-aware, not just file-aware
| Type | Line |
|---|---|
| Review workflow sample | Previously on episode seven, Rina broke the promise. |
| Single-file translation | Sebelumnya di episode tujuh, Rina melanggar janji. |
| Series-aware subtitle (Mitsuko) | Di episode tujuh sebelumnya, Rina mengingkari janjinya. |
A single-file translator has no memory of how earlier episodes phrased things. A project with episode context and agency instructions does.
So "broke the promise" lands as mengingkari janjinya, the series-consistent version, instead of melanggar janji, a one-off literal choice an editor would later flag.
The file reaches editorial review already aligned with the series.
The agency localization flow
- Import client subtitle files and project context. Bring in the subtitle files plus the glossary, character notes, and client style rules as context and custom instructions.
- Translate with instructions for terms, tone, and style. The draft follows your terminology and tone rules instead of guessing. Names, terms, formatting preferences, and client rules come from your instructions.
- Review, QC, and deliver final subtitles. Your translators and QC editors review the draft and own the final sign-off. Mitsuko produces the first pass; the agency produces the deliverable.
Batch processing for recurring client work
Agencies rarely get one file at a time. You get a season, a course, or a recurring weekly drop.
Batch projects let you load multiple subtitle translations at once with shared context. That keeps terminology stable across files without a human re-checking every term decision by hand.
For recurring client work, that is where the time goes and where the consistency risk lives.
Who this is for
- Small and mid-sized localization agencies that want a context-aware first draft, not a black-box auto-publish.
- Subtitle vendors handling recurring files where term consistency across drops is non-negotiable.
- Translator teams that need a consistent, instruction-following draft before QC - so editors spend their time on craft, not on fixing a flat machine pass.
The honest boundary
Mitsuko does not replace translators. It translates subtitles with context and instructions, and your team keeps final review and QC.
If a workflow promises to publish subtitles with zero human review, that is a different product and usually a different quality tier than an agency is paid to deliver.
The value here is a better starting point for the humans who do the real work.
Related workflows
- Batch subtitle translation - multi-file projects with shared context and terminology.
- Audio to subtitles - start from media when clients send files without subtitles.
- ASS subtitle translator - preserve styling and structure in styled deliverables.
- YouTube subtitle translator - creator-facing localization.
- Pricing - how agency-scale usage is billed.
Import the client files, set the glossary and instructions, and move from raw files to a review-ready draft your team can actually build on.