Sinhala speech to text: turning audio into Sinhala subtitles

Last updated: 3 October 2026 · AutocaptionsLK, Sri Lanka

Speech to text for Sinhala is a draft-first job: the machine does the typing and the timecodes, you check the names, numbers and loan words. This page covers what the tool accepts, how to record so the draft is usable, what to correct before publishing — and the cases where typing by hand is still the faster choice.

What the tool actually does with your recording

You upload one audio or video file, the speech is transcribed into Sinhala text, and the lines come back with timecodes attached. You get Unicode SRT and VTT (for editors and players), TXT (a plain transcript) and FM Abhaya copies for old Sinhala software. Nothing is installed on your machine, and the upload is deleted once the captions are ready.

Recording so the transcription comes out clean

• Get close to the microphone. Ten to twenty centimetres from a phone or lapel mic beats a distant room mic every time.
• Keep the room quiet. Fans, traffic, air conditioning and echo all compete with the voice.
• One speaker at a time. Overlapping voices are the hardest case for any speech recogniser.
• Avoid music under the speech. Background music degrades recognition quality even when it sounds faint to you.
• Record the final take. If you re-cut the video afterwards, the subtitles no longer match the audio you sent.

Supported files, size and minute limits

• Audio: MP3, WAV, M4A, AAC, OGG.
• Video: MP4, MOV, MKV, WEBM, AVI — the audio is transcribed, the picture is ignored.
• Limits: one file at a time, up to 150 MB and 15 minutes per file. A lower-resolution export uploads and processes faster.
• Minutes: prepaid, charged by decoded audio duration, rounded up to the next whole second per file.

What to fix in the draft before you publish

1. Names and places. Personal names and village or town names are where a recogniser guesses most.
2. Numbers. Dates, prices, phone numbers and measurements — read them against the audio rather than trusting the text.
3. English loan words. Words that Sinhala speakers often use in English can come back transliterated, in Latin script, or both ways in the same file. Make them consistent.
4. Punctuation and line splits. Break long sentences at natural pauses and keep each caption to one or two lines.
5. The edges. The first and last caption of a file are where timing drift shows up first.

When typing by hand is still the better choice

• Clips under about thirty seconds, where checking a draft takes longer than typing it.
• Heavy specialist vocabulary — medical, legal or technical terms you would have to correct anyway.
• Interviews where several people talk over each other.
• Songs and sung lyrics, which speech recognition handles poorly.

Where the text can go next

• CapCut Desktop or Web — import the Unicode SRT and style the captions (CapCut guide).
• Adobe Premiere Pro — import the SRT as media, then drag it onto the sequence.
• YouTube — upload the SRT as a subtitle track (YouTube guide).
• Documents — the TXT output is the transcript, useful for an article, subtitles-less platforms or accessibility notes.

Sinhala caption packages

Caption minutes are prepaid and used per decoded audio second. Prices are in Sri Lankan rupees and payment is confirmed on WhatsApp before a private activation key is issued.

One-time minute packages, not monthly subscriptions. Usage is based on decoded audio duration, rounded up to a whole second per file. Failed jobs return their reserved credits, while completed jobs use them. Activation keys have no calendar expiry, but availability depends on the service. Purchases are arranged manually on WhatsApp and packages are non-refundable except where required by law — read the refund policy first.

Frequently asked questions