Sinhala subtitle generator

How the Sinhala subtitle generator handles a recording

The Sinhala subtitle generator uses Whisper in the browser, with Sinhala selected in the language picker. The audio is processed in the tab and the transcript returns in Sinhala script. Auto-detect can work for a clean, single-language clip, but choosing Sinhala is more reliable when a video opens with music, English narration or a title card.

Many Sinhala recordings mix English words, names and technical terms into everyday speech. The first pass gives you a timed draft. A Sinhala speaker still needs to check spelling, word boundaries, names and the places where the speaker changed language.

Sinhala script, shaping and left-to-right layout

Sinhala is written in the Sinhala script, an abugida in which consonants carry an inherent vowel and vowel signs, called pili, attach to the consonant. Some combinations form ligatures or change shape, and prenasalised consonants have their own forms. A font with incomplete coverage can show dotted circles or separated vowel signs even when the underlying text is correct.

Unlike Sindhi or Pashto, Sinhala reads from left to right. The editor and the exported cues keep the normal left-to-right flow, so the main risks are font coverage and line breaking rather than base direction. Set the language attribute to si so the browser chooses a suitable font and assistive technology can identify the text.

Sinhala has a relatively large glyph repertoire and tall vowel signs. Allow enough line height and avoid crowding two lines against the bottom of the frame. Test a cue that contains a prenasalised consonant, a vowel sign above the base and a closing punctuation mark, because those features expose clipping and shaping problems quickly.

Punctuation and numerals in Sinhala subtitles

Modern Sinhala text generally uses Western punctuation: the full stop, comma, question mark and exclamation mark. The traditional kunddaliya ෴ is mainly historical or decorative and is not needed in ordinary subtitles. Use the punctuation that matches the register of the video and keep it consistent across cues.

Contemporary Sinhala uses Western digits 0123456789 in everyday digital text. Historical Sinhala numerals exist, but they are not the normal choice for a modern video caption. Dates, phone numbers, prices and scores should therefore be written in the same digit style as the rest of the project, and each number should stay in one run.

Because Sinhala is left to right, numbers do not reorder in the way they can in Arabic-script subtitles. The common problem is different: a line may become too dense when a long number, an English brand name and a Sinhala sentence share the same cue. Split the cue or move the detail into a second line rather than shrinking the font.

Sinhala dialects and spoken variation

Sinhala has regional and social variation, including low-country, up-country and Sabaragamuwa speech, and a noticeable difference between formal written Sinhala and colloquial spoken Sinhala. A transcription of spontaneous speech will often be closer to the colloquial form than to the formal prose found in a textbook.

Do not automatically convert every colloquial form into formal written Sinhala. The subtitle should match the speaker's register. An interview, a comedy clip and a news bulletin can legitimately use different levels of formality, and a reviewer should decide how much normalisation is appropriate.

Spoken Sinhala also includes borrowed words from English, Tamil and Portuguese. Some are fully integrated, while others are code-switched on the spot. Listen for the speaker's actual choice rather than replacing a borrowed word with a formal Sinhala synonym that changes the tone.

Names, honorifics and code-switching with English

Sinhala personal names often include several components, and the order used in speech may differ from the order in an official document. Titles, kinship terms and honorifics can be important to the meaning of a line. Check a person's preferred spelling and the form of address used in the video before you apply a name list across the whole file.

English code-switching is very common in urban and professional Sinhala. A speaker may use an English noun inside a Sinhala sentence or switch to a full English clause. Keep the English in Latin script when that is what was said, and do not force a Sinhala transliteration. This page does not promise romanized Sinhala, sometimes called Singlish, as an output mode.

Place names and organisation names may have an official Sinhala form and a familiar English form. Choose one convention for the project and apply it consistently to the captions, the video title and any description. Mixing conventions within a single cue makes the subtitle look careless even when every individual spelling is defensible.

Line length and timing for Sinhala captions

Sinhala captions read best when they are short. Two lines are easier to follow than three, and a line should break between words rather than inside a consonant-vowel cluster. A reasonable starting point is about forty characters per line, but treat it as a starting point: the script's tall vowel signs and the chosen font can make a line feel much longer.

Keep reading speed under control. If a cue contains a long personal name, a number and a parenthetical remark, split it even when the character count seems acceptable. The viewer needs time to recognise the script as well as read the words. Check the timing in the player at normal speed and on a phone, where the safe area is smaller.

Avoid placing a line so low that descenders or vowel signs touch the edge of the frame. If the video will be burned in, inspect a frame with a tall mark and a low descender together. That is where clipping usually appears first.

Review workflow for Sinhala subtitles

Review the draft in stages so that language problems and presentation problems do not get mixed together. Completeness comes first, then names and numbers, then punctuation and line breaks.

1. Watch the whole clip with the draft visible and note missing speech, merged cues and language switches.
2. Check personal names, place names, dates and numbers against a written source or a Sinhala reviewer.
3. Decide how colloquial or formal the captions should be and apply that register consistently.
4. Preview at normal speed and at phone size, then shorten any line that is hard to read.
5. Export UTF-8 SRT or VTT and test the file in the player that will actually show the video.

Privacy and export for Sinhala work

Whisper runs inside the browser, so the recording does not need to be uploaded to an inference server. Model files are downloaded to the browser and no account is required for the workflow on this page. Privacy still depends on the device and the browser profile: shared computers, synced profiles and extensions can expose local files.

Export SRT for broad compatibility or VTT for formal cue settings. Save the file as UTF-8 so the Sinhala vowel signs and conjuncts survive, then open it in a plain text editor and in the target player. Check the first cue, the longest cue and a cue that mixes Sinhala with an English name.