Frequently asked questions

Lumiox is an AI captioning tool that adds accurate, word-by-word subtitles to Reels, Shorts and YouTube videos in 17 Indian languages. You upload a clip, the spoken audio is transcribed with a start and end time for every word, and you can correct any word before exporting a captioned MP4 or a subtitle file.

Upload your MP4 or MOV, pick the spoken language or let Lumiox detect it, then choose a caption style and export. The whole pass takes about as long as the clip itself. Because the captions are burned into the video on export, they look identical on Instagram, YouTube Shorts and anywhere else you post — rather than relying on each platform's own auto-captions.

No. If you can upload a file and read a transcript, you can caption a video. The editor shows your words as a list you can type over, and the caption styles are presets — you pick one rather than building it. Word-level timing is adjusted by dragging, not by entering timecodes.

MP4, MOV and WebM files from a phone, camera or editor. Vertical, square and landscape all work, and the captions are positioned relative to the frame so a 9:16 reel and a 16:9 YouTube video each get captions sized for their own aspect ratio.

Yes. The free plan gives you 2 minutes of transcription a week on clips up to 1 minute long, with every caption style and language unlocked and no card required. Free exports carry a Lumiox watermark, which you can remove on a single clip for ₹21 or lift entirely by moving to a paid plan.

Lumiox captions 17 Indian languages: Assamese, Bengali, Gujarati, Hindi, Kannada, Kashmiri, Malayalam, Marathi, Nepali, Odia, Punjabi, Sanskrit, Santali, Sindhi, Tamil, Telugu and Urdu. Each can be written in its own native script or romanised into Roman letters, and English is supported throughout.

Yes, and it keeps the sentence together. Most captioning tools treat Hindi and English as separate languages and split a mixed sentence in half, mangling the spelling on either side of the switch. Lumiox transcribes Hindi-English speech as one continuous line, so a sentence that switches language mid-way still reads the way it was spoken.

Yes. Every supported language can be written in its native script or romanised, and you choose per project. Romanised output suits audiences who speak a language but do not read its script comfortably — a large share of younger viewers on short-form platforms.

Accurate enough that correcting a clip is a matter of fixing a handful of words rather than retyping the transcript, but never perfect — no speech model is. Accuracy drops with heavy background noise, several people talking over each other, and uncommon proper nouns such as names and places. That is exactly why every word stays editable before export.

Yes. The model is trained on real-world Indian speech rather than on studio-clean English, so regional accents, everyday slang and casual delivery are the expected input, not an edge case. Uncommon words and names are still worth checking before you export.

No, the language is detected automatically from the audio when you upload. You can override the detection and set the language yourself, which is worth doing when a clip is very short or starts with music before anyone speaks.

Word-level timing means every single word carries its own start and end time, instead of a whole line appearing and disappearing at once. It is what makes a word highlight exactly as it is spoken — the karaoke effect short-form captions rely on — and it is why the emphasis in a Lumiox caption lands on the beat rather than a moment late.

Yes, every word is editable before export. Type over any word in the transcript and the caption updates without disturbing the timing of the words around it. Names, brands and technical terms are the usual corrections.

Yes. Drag a word on the timeline to move it, and only that word moves — the words either side of it stay exactly where they are. Most tools shift everything downstream when you nudge one word, which turns a one-second fix into re-syncing the whole clip.

Yes. Start from any caption style and change the font, the colours, the size and the emphasis treatment. Colour can be set word by word, so the word you stressed out loud can be the only one in your brand colour.

Yes. Caption lines can be split and rejoined so a phrase breaks where it makes sense to read rather than where the transcript happened to wrap. This matters most in the native Indian scripts, where a badly placed break can separate parts of a single word.

Yes, both SRT and VTT are included on every paid plan; the free plan exports captioned MP4s. Use them when you want captions the viewer can switch off, when you are uploading to YouTube as a separate subtitle track, or when you are handing the file to a video editor working in Premiere, DaVinci Resolve or Final Cut.

Burned-in captions are part of the video image and always visible; an SRT is a separate text file the player overlays, which the viewer can turn off. Burn captions in for Reels, Shorts and TikTok, where most people watch on mute and platform subtitle support is inconsistent. Use an SRT for long-form YouTube, where viewers may want to disable them or switch language.

Up to 2K on the paid plans and 1080p on the free plan, and never upscaled — a clip is exported at its own resolution rather than being stretched to hit a headline number. Exports are watermark-free on every paid plan.

Any paid plan exports without a watermark, with no limit on how many clips you export. If you only need one clean file, a one-time watermark removal costs ₹21 and re-opens that clip for export for 24 hours without starting a subscription.

Up to 5 minutes per video on paid plans. Free videos can be up to 1 minute long. Your monthly allowance is measured in total transcription minutes, so a 90-minute plan covers roughly 180 reels at 30 seconds each.

Plans start at ₹299 a month for 90 minutes of transcription, ₹499 for 180 minutes and ₹999 for 480 minutes, with 15% off when billed yearly. There is also a free plan with 2 minutes a week. Every paid plan includes watermark-free exports, SRT and VTT files, and all caption styles.

The duration of the audio you transcribe, rounded to the nearest second — a 40-second reel uses 40 seconds of your allowance. Editing captions, restyling them and re-exporting the same clip cost nothing extra, because you are only charged the first time a clip is transcribed.

No. Only the audio track is uploaded for transcription, and it is deleted once the transcript comes back. The video file itself never leaves your device — captions are rendered and burned in locally in your browser, which is also why export does not wait in a cloud queue.

Captioning pauses until your allowance resets at the start of your next billing month, and nothing you have already made is affected — existing projects stay editable and exportable. Moving up a plan applies the new allowance immediately.

UPI, credit and debit cards, and netbanking, processed through our payment partners Cashfree and Razorpay. You can cancel in one click from your account and keep access until the end of the period you have already paid for. There is no cancellation fee and no notice period.