AI translation is fine… until it breaks your timing, tone, and trust.

TL;DR

  • Multilingual subtitling is not a translation problem, it is a timing, structure, and accuracy problem

  • Always translate from the approved subtitle file, not the script

  • Use AI for fast first pass translations, but only inside the existing SRT structure

  • Use human translators for judgement, nuance, and native speaker credibility

  • Lock the edit and subtitles first, then translate, review, and sense check against the video

“Let’s translate the transcript for a global audience”

If you think multilingual subtitling is just running your script through Google Translate, you are already shipping bad content.

Subtitles live inside time, not text.

That is why multilingual videos often feel “off” overseas.
The words might be translated, but the meaning, pacing, and credibility are not.

The fastest way to undermine an international campaign is to treat subtitling like an afterthought.
One mistranslated subtitle can undo months of brand work.

Stop translating your script.
Start subtitling properly.

We have shared how to do this using subtitle files, AI, human translators, or both, without needing to speak the language yourself.

Example from Australian Agtech:

In our work with Bitwise Agronomy, an Agtech company producing insight-led video content targeting South America, this assumption surfaced early. The challenge was not translating the words themselves, but preserving meaning, accuracy, and pacing in technical material. By working directly from the approved English subtitle file, we were able to create translated subtitles that stayed aligned with the original edit, without the client (and us) needing to speak the target language.

Why we do not rely on platforms to translate subtitles

Many platforms now offer automatic subtitle translation. For professional use and representing brands and important messaging, this is rarely enough.

The main issues are simple:

  • Accuracy: Platform translations have a higher error rate and no human accountability. Small mistakes in wording, tone, or terminology matter when representing a brand or organisation to native speakers.
  • Brand and message control: Subtitles shape how your message is received. Platform translations make silent editorial decisions that you cannot review, approve, or stand behind.
  • No flexibility for real production use: Subtitles are often hard-coded into videos for social media, styled as branded subtitles, or used as a visual design element in the edit suite. Platform translations sit outside the edit and cannot guarantee timing, line breaks, or layout.

For these reasons, we treat subtitle translation as part of the production process, not a playback feature added afterwards. Platforms are improving, but they are not yet reliable enough for professional multilingual delivery.

Start with what is approved

Before any translation begins, three things must be locked:

  • The final approved video
  • The final approved English subtitles
  • A clean export of the subtitle file

This is not bureaucracy. It prevents errors from compounding later.

If the video timing is changing, the subtitles will shift. If English subtitles are still evolving, any translation will either need to be redone or will drift out of sync with the video.

We most commonly work with SubRip subtitle files, known as SRT files, but the same principles apply to formats such as VTT, STL, ASS, SCC, and XML-based subtitle files used by broadcasters and platforms.

Why the subtitle file matters

A subtitle file shows three things that a script does not:

  1. Subtitle numbering
  2. In and out timecodes
  3. Exactly what appears on screen

When this structure is preserved:

  • Each translated sentence appears at the correct moment
  • Line breaks remain readable
  • The translated language stays aligned with speech

This is what allows an English speaker to manage multilingual subtitles without speaking the target language.

Example from German Engineering:

On projects for thyssenkrupp, the original language of the interviews recorded was German.

We first transcribed the German audio, translated it into English, and used that English translation to drive the edit. Only once the picture was locked did we then create structured subtitle files in both German and English.

This ensured editorial decisions were driven by meaning – not guesswork, both language versions were structurally aligned and subtitles in each language were accurate and credible to native speakers.

Translations were reviewed line by line by native German speakers, working directly from the subtitle files to ensure terminology, sentence structure, and timing reflected professional engineering usage.

Using AI properly

AI translation tools can be extremely useful for subtitles when they are used with clear constraints. When using AI to translate an SRT file, the most important instruction is this:

“Only translate the text inside each numbered subtitle block. Do not change numbering, timecodes, line breaks, or structure.”

AI should operate inside the existing file, not alongside it.

Used this way, AI can provide fast, cost effective first pass translations. It cannot reliably judge how a translation will feel to a native reader watching a video, and it should not be asked to make editorial decisions.

Working with human translators

When working with human translators, we use a different but equally simple instruction set. The aim here is not speed, but judgement. We convert the .srt to a .txt file and provide the following simple instructions:

INSTRUCTIONS: Do not remove any of the number or timecode formatting. Please follow the example below for your translation. If any formatting of the text, line 1 to line 2, does not work for the translation, please note this change.

This structure allows translators to work line by line, preserving timing while flagging issues that need review rather than silently fixing them. Any adjustments that need to be made due to sentence structure or timing can then be matched to the English subtitle.

Combining AI and human review

In practice, the most robust approach is often a combination of both.

A typical workflow might look like this:

English subtitle file approved > AI used for a first pass translation > Review and refinement by a bilingual or native speaker > Line by line checks against timing > Final sense check against the video

This balances speed, cost, and accuracy without compromising trust.

Here is an example of a simple translation format using the original .srt format converted to a text file and following the instruction above.

ORIGINAL:

1

00:00:00,639 –> 00:00:02,719

<b>The Edinburgh International</b>

<b>Book Festival</b>

TRANSLATION:

1

00:00:00,639 –> 00:00:02,719

<b>爱丁堡国际</b>

<b>图书节</b>

Example from Scottish Event Tourism:

On projects for the Edinburgh International Book Festival, the original language of the film was English, but the audience was international, and the marketing was targeted at specific countries.

We first locked the final edit and approved English subtitles, then translated the subtitle file into German, Mandarin, and Spanish, working directly from the timed SRT rather than the script. This ensured editorial decisions remained consistent across all versions, the same film could be reused without re-editing, and each language version stayed structurally aligned to the original timing.

Translations were reviewed line by line by native speakers, working directly from the subtitle files to ensure meaning, reading speed, and sentence structure felt natural to audiences watching in their own language. We then produced both a landscape 16:9 and a vertical 9:16 version.

SIDEBAR: Subtitles and closed captions are not the same thing

It is worth pausing here to clarify a distinction that is often overlooked.

Subtitles are designed to translate or support the understanding of spoken dialogue. They assume the viewer can hear the audio, and they focus on what is being said. Subtitles do not usually describe background sounds, music, or nonverbal audio cues.

Closed captions, on the other hand, are an accessibility feature. They are intended for viewers who cannot hear the audio or cannot hear it clearly. In addition to dialogue, closed captions include information such as:

  • Sound effects
  • Music cues
  • Speaker identification
  • Tone or delivery where relevant

This distinction matters for multilingual work.

If you are translating subtitles into other languages, you are typically translating dialogue-only. If you are producing closed captions, especially for accessibility compliance, those captions need additional care, structure, and often a different editorial approach.

In many projects, both are required. Understanding which one you are creating, and why, should come before any translation begins.

The simple truth

Creating multilingual subtitles is not complicated, but it is precise. Most problems arise from translating the wrong source, skipping steps, or breaking the relationship between language and time.

With a locked video, approved subtitle file, clear structure, careful use of AI and human review where it matters

You can create accurate, respectful, and effective multilingual subtitles without speaking the language yourself.

If you would like to see the subtitle translation template we use, or talk through the right approach for your organisation, we are always happy to help.

Subtitles are designed to translate or support the understanding of spoken dialogue. They assume the viewer can hear the audio, and they focus on what is being said. Subtitles do not usually describe background sounds, music, or nonverbal audio cues.

Closed captions, on the other hand, are an accessibility feature. They are intended for viewers who cannot hear the audio or cannot hear it clearly. In addition to dialogue, closed captions include information such as:

  • Sound effects
  • Music cues
  • Speaker identification
  • Tone or delivery where relevant

This distinction matters for multilingual work.

If you are translating subtitles into other languages, you are typically translating dialogue only. If you are producing closed captions, especially for accessibility compliance, those captions need additional care, structure, and often a different editorial approach.

In many projects, both are required. Understanding which one you are creating, and why, should come before any translation begins.

Looking for a content partner?

Book a short discovery call

A short call to get to know eachother – no hard sell. All we’ll do is aim to understand your goal, recommend the right format, and agree next steps.

No formal prep needed. We use Google Meet and capture a shared transcript for accurate notes. No video or audio is stored.

Explore more insights…

Turning social viewers into website visitors

Turning social viewers into website visitors

Turning social viewers into website visitors TL;DR When deliverables feel fuzzy, start with a flagship video plus a social trailer. It is a dependable pairing that fits most objectives. Social trailers earn attention in feed and send qualified viewers to your flagship...

read more