Sri Lanka is a small market on a world map and a difficult one to serve well. Sinhala is the majority language and an official language of the state, spoken by roughly 22 million people on the island plus diaspora communities in the Gulf, the United Kingdom, Australia, Canada, and Italy. If your library is in English, Spanish, Hindi, or Korean, there is a credible Sinhala audience for it, and almost nothing in a standard pipeline produces Sinhala output by default.
Treating Sinhala as one more line item in a bulk translation order fails for predictable reasons. It has a sharp gap between literary and spoken registers that decides whether a dub sounds like a news broadcast or a conversation. It uses an abugida whose combining vowel signs sit above, below, and beside consonants, which changes what a subtitle line can physically hold. And text length against the soundtrack rarely behaves as translators expect, so timing drifts in ways a non-speaker will not catch by watching once.
This guide covers the whole path for a creator who needs to translate video to Sinhala from an existing library: what to decide before the first run, how to produce subtitles and audio, what belongs in a glossary, how voice cloning handles a recurring presenter, and how to review output you cannot read. The route from an English master to a finished Sinhala release runs through video translation, but the decisions after that point are where projects fail. Review is the weakest link, and a structure you can check without speaking Sinhala separates a published video from an embarrassment.
Who the Sinhala audience is, and why Tamil usually travels with it
The audience is bilingual in ways that change your release plan
Sri Lanka's language situation is not a single-language market with a minority. Sinhala and Tamil are both official languages, and much of the Sinhala-speaking audience watches English-language content with English or Sinhala subtitles. Localizing for Sri Lanka rather than the diaspora alone usually means two tracks competing for the same review budget.
Diaspora audiences often want Sinhala audio with English subtitles; domestic audiences want Sinhala audio and Sinhala subtitles, or Sinhala subtitles over the original audio. Decide which of the three you are shipping before the first render, because the answer changes the deliverable list.
- Sinhala dub plus Sinhala subtitles: domestic viewing on phones, where sound is usually on but subtitles help in noisy environments.
- Sinhala dub plus English subtitles: diaspora households where the second generation reads English more comfortably than Sinhala script.
- Sinhala subtitles over the original audio: the cheapest option, reaching viewers who already tolerate reading, which skews younger and more urban.
Where the language actually travels
Sinhala is not widely understood outside Sri Lanka and its diaspora, so the return on a Sinhala dub comes from depth in one audience, not breadth across many. A Tamil track doubles the number of people who can consume the localized version, but it is a separate production with its own reviewer, glossary, and register decisions.
The register split that decides whether your dub sounds right
Literary and spoken Sinhala are far apart
Written formal Sinhala and conversational spoken Sinhala differ more than the equivalent gap in English: verb endings, pronouns, and vocabulary all shift. A model working from text tends toward the literary register, which reads fine on a subtitle line and sounds stiff in a voice track. A conversational interview dubbed in literary Sinhala sounds like a government announcement.
Decide the register before the first run and write it into the instructions. Two options cover most cases.
- Spoken-conversational for dubbing: matches a presenter's natural speech, uses contracted verb forms, and avoids formal pronouns unless the source genuinely uses them.
- Neutral-formal for subtitles: closer to broadcast written Sinhala, more stable across lines, easier to read quickly.
The trap is assuming one translation serves both. Subtitle text that reads well is often too dense to speak at pace, and dubbed speech transcribed verbatim makes clumsy subtitle lines.
Formality is a translation decision you have to make explicitly
Sinhala marks respect through pronoun and verb choice. A second-person address that is neutral in English has to be resolved into a formality level, and the wrong choice reads as rude or oddly distant. This matters most in instructional content, product videos, and anything that addresses the viewer directly.
Specify it up front rather than letting each translator guess: use the polite but not deferential second person, as if addressing an adult customer you do not know; keep technical vocabulary in its commonly borrowed English form where the Sinhala term would be unfamiliar; avoid honorific forms reserved for elders and religious contexts.
Mixed code-switching is normal and should be preserved
Sinhala speech in technical and business contexts routinely borrows English nouns mid-sentence. Over-purifying produces text that reads as archaic, and a dub that avoids all English terms sounds unnatural to listeners who use them daily. For software content, specify that common English technical terms stay in English while Sinhala carries the grammar.
How the Sinhala script constrains subtitles
The abugida changes line height, not just line length
Sinhala is written in an abugida: consonants carry an inherent vowel, and other vowels are marked with combining signs placed above, below, before, and after the consonant. Those marks are part of the character, not decoration. A consonant cluster can stack a vowel sign above the letter and another below it, so glyphs run tall. When a font lacks full coverage, rendering fails as broken or disconnected marks rather than a visible box.
- Use a font with verified Sinhala coverage and correct shaping, and test it with stacked vowel signs rather than a simple word.
- Allow more line height than Latin subtitles need, so vowel marks are not clipped.
- Avoid thin or condensed weights, which lose the distinction between similar marks at video resolution.
- Check rendering on the smallest target device, not on a desktop preview where everything looks fine.
Line length and reading speed need different numbers than Latin text
Sinhala words are often longer than their English equivalents when written out, because suffixes and verb forms carry grammatical information English handles through word order, so lines run wider for the same meaning. Treat Sinhala subtitles as you would a long-word European language: fewer characters per line than your English baseline, and slightly more time per line for an audience reading the script less fluently than they speak it.
A defensible starting point is a maximum of roughly 42 Latin characters or the equivalent visual width, two lines at most. Verify by watching the render at phone size and pausing on the densest lines; if a line requires squinting, it is too long regardless of character count. If your library already has captions, review what a subtitle generation pass produces before you translate, because a clean source track makes every downstream step cheaper.
When text expansion or contraction breaks the timing
Sinhala against English does not expand uniformly. Some phrases shrink because Sinhala packs meaning into inflected verb forms; others grow because a technical term needs a phrase rather than a word. Problem cues cluster rather than distribute evenly, so the costly sections are rarely the ones you would guess. Working from a subtitle translation that preserves the original cue structure gives you a stable base for the timing pass.
When a line no longer fits its cue, there are four options and only two of them are good.
- Rewrite the translation shorter, dropping redundancy while preserving meaning. This is almost always the right first move.
- Rebalance timing across neighboring cues that have slack.
- Increase reading speed. Acceptable in short bursts, a defect if it happens throughout.
- Split the line across two cues. Sometimes necessary for a long sentence, but it disrupts rhythm and is the last resort.
Do not shorten display duration below the comfortable reading threshold or let cues overlap. Both produce the same complaint: the subtitles flash by.
What to decide before you translate video to Sinhala
Build a glossary first, not after the first review
The most reliable way to avoid a second full pass is a glossary produced before the first translation run. Translators working independently make different choices for the same term, and inconsistent terminology is the defect viewers notice fastest.
The glossary should cover:
- Product and brand names: which stay in Latin script, which are transliterated, which are spoken in English in the dub.
- Personal and place names: one transliteration spelling, fixed rather than decided per translator.
- Recurring technical terms: the chosen Sinhala term, or the decision to keep the English word.
- Units, currency, and measurements: convert or keep, and how to write them.
- Numbers and dates: Sri Lankan date order, digit grouping, and whether to spell out numbers in the dub where pronunciation is ambiguous.
- Honorifics and titles: how names that carry a title in the source are handled.
The same glossary applies to the subtitle file, the dubbing script, and any text burned into the video. A term transliterated one way on screen and pronounced another in the audio is a visible defect.
Transliteration is a choice, and consistency matters more than the choice
No single correct way to render English names in Sinhala script exists, and established conventions differ. What matters is that one convention holds throughout the deliverable and matches how the name is pronounced in the dub. Fix it in the glossary with a written example and give that glossary to everyone who touches the project.
Producing the audio: voice selection and cloning
Matching the source voice is a decision, not a default
Two viable approaches exist: a different speaker whose voice fits the content, or synthesis that preserves the original presenter's voice across languages. The first is simpler and, for a one-off video, entirely reasonable. The second matters when the presenter is the brand: a recurring host, a course instructor, a founder who appears in every episode. Replacing that voice breaks continuity viewers notice. Video dubbing puts the new performance against the original timing, which is what keeps the result from drifting.
Voice cloning is a shipped, working capability, and preserving a speaker's voice across a translation is what it is for. Two conditions apply, both operational rather than technical. You need written authorization from the speaker whose voice is cloned, covering the languages and intended use. And you should disclose that the voice is synthesized, in the description or a closing card, in whatever way fits your channel's norms and any applicable platform rules. If the source audio is the only reference, start from a clean, dry sample; a clone built from a track with music under the speech carries an audible artifact of the music.
Handling overlap, music, and multiple speakers
Dubbing a conversation is harder than dubbing a monologue because of overlap. If two speakers talk over each other, a straight replacement collides in the dub. Speaker separation and dialogue detection handle most of this by isolating dialogue from the music bed and keeping speaker turns distinct, and the surviving music can be laid back under the new speech. Where overlap carries meaning, expect a small timing adjustment so the exchange stays legible in Sinhala. A single presenter to camera is far easier and is the right place to start.
A quality-review checklist a non-speaker can run
You cannot judge Sinhala fluency. You can judge everything mechanical, and most published failures are mechanical. Work through this in order; later checks are meaningless if earlier ones fail.
- Render check: no missing glyphs, no disconnected vowel signs, no clipped marks at the top or bottom of the subtitle area.
- Coverage check: every spoken segment has a cue and every cue has audio, with no gaps at scene transitions.
- Timing check: pause on the five densest lines and confirm each is readable at normal speed on a phone.
- Sync check: audio enters and exits with the on-screen mouth movement, allowing for Sinhala sentence structure that may reorder the line.
- Name and number check: every glossary term appears as specified in both the subtitles and the audio.
- Truncation check: line breaks do not split a word or separate a vowel sign from its consonant.
- Duration check: watch the full video once at normal speed without pausing, the only way to catch cumulative drift.
Then the native review. Brief the reviewer on specific things rather than asking whether it is good. A useful brief asks: does the register match the source tone; are there lines that sound like machine output; is any term wrong or offensive; does the direct address strike the right formality; do the names sound like the names they are meant to be. Ask for a timestamped list of specific lines and a corrected version of the ones that are wrong.
A back-translation into English surfaces gross errors, dropped sentences, and terminology drift, but not register or naturalness problems, so treat it as a filter rather than a substitute. For reviewer handoff, keep the source script, the translated script, the glossary, and the subtitle file together, and give the reviewer the video with timecode visible. Subtitles export as SRT or VTT, so the file the reviewer corrects is the file that ships; the documentation covers supported formats and export options.
How to translate video to Sinhala: the first-project plan
Start with the videos whose failure is cheapest
Do not begin with your flagship. Begin with a small set that makes localization easy to evaluate.
- A single-presenter explainer under ten minutes, with continuous speech and little overlap.
- Content that is already evergreen, so a delay costs nothing.
- A piece with a companion transcript, so a clean subtitle pass is quick.
- Something whose audience you can measure, ideally on a channel or playlist you can compare against a baseline.
Exclude anything with heavy on-screen text baked into the video, rapid multi-speaker debate, or humor that depends on wordplay. Those are worth doing once the glossary and the reviewer relationship exist.
Measure whether it worked
Compare per view rather than per video.
- Watch time and average view duration on the Sinhala version against the source version.
- Retention at the first thirty seconds, where bad audio or slow-reading subtitles cost you most.
- Completion rate, which separates a translation that is fine from a dub that is exhausting to listen to.
- Comment content, read for specific complaints about audio, terminology, or subtitle speed rather than for sentiment.
- Whether viewers finish and then watch a second video in the same language, the strongest signal that the dub is tolerable.
High click-through with low retention usually means the thumbnail and title translated well and the audio did not, which points at voice and register. Even retention with lower click-through points at discoverability instead, a metadata problem rather than a translation problem.
Sequence the work so the second project is cheaper
The order that pays off: fix the glossary, decide the register, produce one clean subtitle track, translate it, generate the audio, review mechanically, then send it to a native reviewer. A locked subtitle track doubles as the dubbing script, which is what makes subtitle-to-audio conversion practical. Every artifact is reusable: the glossary carries forward, and the reviewer's corrections become glossary entries. By the third video the marginal cost per minute drops sharply, which is the reason to build the process rather than buy a one-off translation.
Frequently asked questions
Can Sinhala subtitles be generated automatically from the audio?
Yes. Generation works from the audio track and produces a timed transcript, which you translate rather than re-time. With a well-supported source language you get accurate timings, and the translation step then works on text.
Should the dub and the subtitles use the same translation?
Usually not. The dub needs spoken register and natural pacing; the subtitles need compact, readable lines. Start from one translation, then edit each for its medium.
How do I check quality if nobody on my team speaks Sinhala?
Run the mechanical checks yourself: rendering, glyph integrity, timing, coverage, glossary consistency, and sync. Those catch most published defects. Then hire a native reviewer with a brief that asks for timestamped problems and corrected lines.
What should the glossary contain before the first run?
Product and brand names with their treatment, personal and place name transliterations fixed to one spelling, recurring technical terms, units and currency handling, number and date formatting, and honorifics.
Is voice cloning necessary for a Sinhala dub?
Only when the presenter's voice is part of the content's identity, such as a recurring host whose absence would be jarring. For standalone videos, a natural speaker who fits the material is simpler. If you clone, get written authorization and disclose the synthesized voice.
How long should the first Sinhala project take?
Plan it as a learning project. The glossary and register decisions consume the early effort, translation and audio are comparable to other languages, and the review depends on a native speaker's calendar. Budget one correction pass.
Do I need a separate Tamil version?
Only if you are targeting Sri Lanka domestically rather than the Sinhala-speaking audience alone. Tamil is a distinct language with its own reviewer and glossary. For a diaspora audience, Sinhala with English subtitles covers more of them.
Conclusion
Start by deciding what the Sinhala version is for: a domestic Sinhala audience, a diaspora audience that may read English more comfortably, or a Sri Lanka-wide release that needs a Tamil track alongside. That decision fixes your deliverable list before any translation begins, and it is expensive to reverse later.
Then build the two artifacts that make everything else repeatable: a glossary and a register decision. Fix how names, technical terms, units, and dates are handled once, and state whether the dub is spoken-conversational and the subtitles neutral-formal. With those in hand, the translation and audio steps become routine, and the review becomes a checklist rather than an argument.
For the first project, pick one evergreen single-presenter video under ten minutes with an existing transcript, produce the subtitle track first, translate and dub from it, run the mechanical checks, then send it to a native reviewer with a specific brief. Compare retention and completion against the source version. If the numbers hold and the reviewer's corrections are mostly terminology, the process works and the next video is cheaper. If retention collapses in the first thirty seconds, the problem is almost always audio and register, and that is where the next iteration should go.
