Most free caption tools handle English well and Bangla badly. They either write your বাংলা in the wrong alphabet, spell it out in English letters, or give up and repeat the same few words. If you make videos in Bangla for Facebook, Reels, TikTok or YouTube, you have probably tried a tool, seen the result, and gone back to typing every line by hand.

This guide shows how to add Bangla subtitles to a video automatically on your iPhone or Mac, how accurate to expect it to be, and how to fix the lines it gets wrong. Everything runs on your device, and your video is never uploaded.

Why Bangla is hard for automatic captions

Speech models are trained on far more English than Bangla. A general-purpose model hears Bangla, recognises that it is not English, and then has to choose a writing system. Many choose the wrong one: Arabic-style Urdu letters, or Latin letters that spell the sounds. We tested the general models Zella used before 1.0.7 and they scored about 100% character error on a plain Bangla sentence, which means essentially none of it was right.

A video frame with captions burned in, one word highlighted as it is spoken.

What fixed it is a Bangla-specific model: a speech model fine-tuned on Bangla recordings, then converted to run on a phone or Mac. It writes বাংলা in Bangla script, which is the part general models could not do.

Apple's Speech framework shows that speech can be transcribed on the device, with no upload. Zella transcribes with its own captions model, which ships inside the app from version 1.0.7, so the recording stays where it was made.

Treat captions as access first and engagement second. The standard most teams are held to is the W3C's captions for prerecorded media.

What you need

  • Zella 1.0.7 or later, on iPhone (iOS 18 or later) or on Mac (macOS 14 or later).
  • A video where the speech is mostly Bangla.
  • About 500 MB of free space for the Bangla model, downloaded once.

If you are on an earlier version, update first. Zella 1.0.6 and older do not caption Bangla.

How to add Bangla subtitles on iPhone

  1. Open your project and tap Captions in the tool row at the bottom. The Studio sheet opens on its Captions tab.
  2. Set Spoken language to Bengali / বাংলা.
  3. Tap Generate captions.
  4. A pop-up offers the free Bangla captions model. Tap Download free. Zella shows the size first, about 485 MB, and downloads it from zellahq.com.
  5. When the download finishes, captions generate on their own. They appear in Bangla script on your timeline.

From the second video on, there is no pop-up and no waiting: the model is already on your phone.

How to add Bangla subtitles on Mac

  1. Open the video in Zella's editor and click the Captions tab on the right.
  2. Pick Bengali / বাংলা in the language menu beside Transcribe audio.
  3. Click Transcribe audio and accept the free Bangla model when asked. It is about 539 MB and downloads once.
  4. Pick a caption style and export as usual.

You can also leave the language on Auto-detect. Zella listens to the start of your take, hears Bangla, and switches to the Bangla model by itself.

How accurate is it?

We measured it on a recorded Bangla sentence read by a native voice and scored character error: the share of characters that were wrong, missing or extra. Lower is better.

Where Model Character error
Mac Bangla model (larger) about 5%
iPhone Bangla model (smaller) about 20%
Either, before 1.0.7 General models about 100%

In plain terms, the Mac model gets roughly nineteen of every twenty characters right on a clear sentence, and the iPhone model roughly four of every five. The iPhone model is smaller because the larger one took far too long to load on phone hardware. Both write real Bangla script.

One sentence is a quick check and not a promise. Fast speech, background music, a noisy room or a lot of English mixed in will all cost accuracy, in every tool.

How to fix the lines it gets wrong

Read the captions once before you export. On iPhone, tap any caption on the canvas to edit its words; the timing stays. On Mac, edit the transcript like text, and deletes ripple the video.

Two habits make editing quicker:

  • Record close to the microphone. A clear voice does more for accuracy than any setting.
  • Speak in short sentences. Long run-ons give the model more chances to lose the thread.

Which font to use for Bangla captions

Zella's built-in caption fonts such as Montserrat and Anton are Latin-only, so Bangla falls back to the system font. For a branded look, import your own Bangla font (.ttf, .otf or .ttc) and use it for captions. See custom fonts in any script.

Does it work offline and keep my video private?

Yes to both. Once the model is downloaded, captioning needs no internet. The only network request is the one-time model file, which carries nothing from your project. Your video and audio never leave your device.

Other languages on the same update

Zella 1.0.7 also brings a captions model inside the app, so captions work offline from the first launch in English, Spanish, Portuguese, French, German, Chinese, Japanese, Korean, Arabic, Russian, Indonesian and more. Hindi and its neighbours get a free bigger model, the same way Bangla does. The full list, with measured accuracy, is in caption languages and models.

Frequently asked questions

Can Zella caption Bangla automatically? Yes, from version 1.0.7, on both iPhone and Mac. It writes in Bangla script and runs on your device.

Is the Bangla model free? Yes. It is free for every user on every plan, downloaded once from zellahq.com.

Does it work without internet? After the one-time model download, yes.

Why is the iPhone version less accurate than the Mac? The iPhone uses a smaller Bangla model because the larger one is too slow to load on a phone. Both write Bangla script; the Mac model makes fewer mistakes.

Can it caption Bangla and English in the same video? Pick Bangla as the spoken language and English words inside your Bangla sentences are usually kept, but a video that switches language for long stretches is better captioned in two passes.

Can I get English subtitles from Bangla speech? On iPhone, yes: turn on English captions (translate) in Studio. See captions on iPhone.

Does Zella caption Telugu? Not yet. It was measured and does not work reliably on-device.