For Tamil Creators
Tamil captions that
read correctly
Paste a Tamil video and get vertical clips whose captions a Tamil reader can actually read. Of the twelve writing systems we tested, Tamil is the one most damaged by the shaping shortcut most video tools take — because ெ, ே and ை are all stored after their consonant and must be drawn before it.
Free tier included — no credit card required
Try it on your video — no signup
Paste a URL, see your viral moments
The worst-affected script we measured
We rendered the same sentence with and without proper shaping across twelve writing systems and compared the pixels. Tamil diverged the most of any of them — 19 times more than a Latin control line, which shifts only by kerning. Every pre-base vowel sign in the sentence lands in the wrong place without HarfBuzz.
Not a font problem
The usual advice is to install a font with Tamil coverage. Necessary, and completely insufficient: the glyphs are present and correctly drawn, they are simply in the wrong order. A font supplies shapes; deciding that ெ is painted before the letter it follows is a separate step.
It clips, reframes and captions
Nova reads the transcript to find the strongest 25-50 second moments, crops them vertical, writes the hook, and burns word-timed captions — the whole clip, not just a subtitle track.
From upload to posted in three steps
1.Upload or paste
Drop a file or paste a link. Nova handles download and conversion.
2.AI finds the moments
Transcription + moment detection rank and cut your most viral segments.
3.Download & post
Vertical, captioned clips ready for TikTok, Reels, and Shorts.
Ready for every platform
Questions? Answers.
Why do Tamil captions come out wrong in other tools?
FFmpeg's `subtitles` filter — the standard way to burn captions and the one most tooling is built on — exposes no text-shaping option, so the renderer falls back to simple shaping: bidirectional reordering and Arabic joining, but no HarfBuzz. Tamil needs HarfBuzz, because ெ ே ை are pre-base vowel signs stored after their consonant and drawn before it. It was reported to libass as issue #682 in February 2023, closed as 'not our bug' since the option belongs to FFmpeg, and has never been fixed. Check your own output: find a word with ெ and see which side of the letter it sits on.
How bad is it really?
On our measurement Tamil showed the largest divergence of any script we tested — 0.740 against a Latin control of 0.038. Latin moves only by kerning; Tamil moves by whole glyphs. It is not a subtle rendering difference, it is words spelled wrong.
Does Nova transcribe Tamil?
Yes, natively, and you can force the language if a code-switched video is being misdetected as English. Captions are word-timed so the highlight follows the speech.
More ways to use Nova
Guides & resources