Kazakh letters in subtitles: why ә, ғ, қ, ң, ө, ұ, ү, һ, і disappear — and how to check your own clip
What happens
You send a Kazakh recording to be cut, get your clips back — and the subtitles read «жа дайымда» instead of «жағдайымда», «к н» instead of «күн». The letters were not replaced by boxes and nothing was flagged. They simply vanished, leaving a gap, and at a glance the line still looks like a normal line.
This is not a speech recognition error. The speech was transcribed correctly, and the subtitle file has every letter in place. They are lost at the last step — when the text is drawn into the frame.
Why it happens
Text in a frame is drawn by a font, and a font is a file with a finite set of characters in it. Cyrillic is almost certainly there: everyone needs Russian letters. But the nine Kazakh ones — ә, ғ, қ, ң, ө, ұ, ү, һ, і — live in a separate Unicode block, and most fonts do not carry them.
Then comes the worst part: nothing reports an error. The subtitle renderer, finding no glyph, draws an empty box or nothing at all. No warning, no red line in a log. The service considers the job done, you get thirty clips with holes in them, and you hear about it from your viewers.
Hence a rule worth knowing before you order anything: the font must be chosen for the letters that will end up in the frame, not for the language you picked in a form. If the recording is Kazakh and the style you selected is a «Russian» one, the letters will be taken from a Russian font — and nine of them will not be there.
How to check your clip in a minute
No software required, and it works on any service.
- Take a line with the rare letters. «Өмірде әрқашан қиындық болады» does the job on its own;
- find it in a finished clip and pause on that subtitle;
- compare it with the subtitle file the service gives you (ours is .srt). The file always has every letter — the damage lives only in the frame;
- look for gaps, not boxes. A missing letter leaves a hole inside a word: «ұстаз» becomes «стаз», «қыз» becomes «ыз»;
- check every style, if the service offers a choice: it is common for one template to render everything and the rest to fail.
How we do it
We answer this with machinery, not with a promise.
- the font is chosen by the letters in the frame — not by the language stated in the order and not by the language of the speech: if Kazakh text ends up on screen, the font is picked for it;
- coverage is verified against the font file itself. We read the font's character table and ask it about each of the nine letters — not «Cyrillic must be in there somewhere», but a direct answer per character;
- a font without those letters never enters the set. The check runs on every subtitle style we ship, and it is mandatory: a template that cannot draw ә never reaches a customer.
What this means for your order
Nothing to configure. Upload a Kazakh recording and the subtitles come out with Kazakh letters in any style. We recognise speech in 10 languages, Kazakh included.
If you already have clips with holes in the words — from another service or from us — they do not need to be retyped: the problem is in the font, not in the text, and it is fixed by rebuilding the clip with a style that carries the letters.