Transcripts
Turn a video into text you can actually use.
A clean transcript in English, Hindi or Hinglish — corrected by you before you download it, and no upload of the video itself.
LumaCaption converts video to text by transcribing the audio with word-level timing in English, Hindi or Hinglish, letting you correct the transcript, then downloading it as plain text, SRT or WebVTT — without the video file ever being uploaded.
Most video-to-text tools hand you a wall of text and a copy button. That is fine if you wanted a rough record and unhelpful if you wanted something to work with, because the two things that make a transcript useful — knowing when each word was said, and the names being spelled right — are exactly what a wall of text throws away.
The transcript here is word-timed and editable. You can click a word to jump the video to it, fix a spelling in place, and download the result as plain text for show notes, as an SRT for a platform that reads subtitle files, or as a WebVTT for a web player.
Three languages, done properly
English, Hindi in real Devanagari, and Hinglish — Hindi-English mixed speech written in Roman letters. That is the whole list, and each one is tuned rather than inherited.
Every word carries its timing
Click a word to move the video there. Useful for pulling quotes, finding a moment, and cutting clips at the right frame.
Correct it before you keep it
Names, products, acronyms and jargon are editable in place, and the correction flows into every format you download afterwards.
Three outputs, free
Plain-text transcript, SRT and WebVTT — unlimited on every plan, including free.
Two engines, and when the difference shows
There are two transcription engines here rather than one. Candy Lite is the generous one and Candy Ultra is the accurate one, and the gap between them is not uniform — on clean, close-miked English about ordinary subjects they land in much the same place.
The gap opens on hard audio: two people talking at once, a room with reverb, a phone recording, strong accents, dense proper nouns, and above all code-switching. A sentence that starts in Hindi, takes an English noun and returns to Hindi is completely ordinary speech in India and an edge case for a general model, which will usually commit to one language and damage the rest of the sentence.
- Candy Lite
- 20 minutes a month on Free, 10 hours on Creator, 50 on Pro. Right for clean solo recordings.
- Candy Ultra
- 5 minutes a month on Free, 3 hours on Creator, 20 on Pro. Right for crosstalk, accents, noise and mixed Hindi-English.
- Hinglish is Ultra
- Producing Roman-script Hindi-English is an Ultra capability. It is listed on the free plan so you can see it, with the minutes gate shown honestly.
- Metering
- Charged per second of audio transcribed. Exports and subtitle downloads are free and unlimited on every plan.
What a transcript is worth beyond captions
The obvious use is subtitles. The less obvious ones are usually worth more: a transcript is the fastest way to turn one recording into a blog post, a newsletter, a set of quote cards, show notes, a searchable archive, or the script for a translated version.
It is also how you find things. A ninety-minute interview is unsearchable as video and trivially searchable as text — and because every word carries its timing, finding the sentence in the text puts you at the right second in the video.
How it works
Drop the video in
Up to 2 minutes free, 30 on Creator, 90 on Pro. The file never leaves your device.
Pick the spoken language
English, Hindi or Hinglish, chosen before transcription so the right engine and script are used.
Correct and download
Fix names and jargon, then take plain text, SRT or WebVTT — free and unlimited on every plan.
Common questions
What languages can it transcribe?
English, Hindi and Hinglish. Hindi comes back in real Devanagari; Hinglish is Hindi-English mixed speech written in Roman letters, which is a separate mode rather than a transliteration of the Hindi output. That is deliberately the whole list — nothing is offered until it has been benchmarked properly.
How accurate is it?
It depends on the audio and the engine. Candy Lite is reliable on clean, close-miked speech; Candy Ultra is the one to use for crosstalk, room noise, strong accents and Hindi-English code-switching, where the difference is largest. Every word is editable afterwards either way, and corrections carry into every format you download.
Can I get timestamps in the transcript?
Every word carries its own timing internally, so clicking a word jumps the video to it. The plain-text download is the words without timings; if you want timings in the file, download the SRT or WebVTT instead — both are free and unlimited.
Does my video get uploaded?
No. The video file stays on your device; only the audio is used to work out the words. That is the difference that lets people transcribe client work, unreleased footage and internal recordings here.
