😈 Caption Videos on Your iPhone Without Uploading Them Anywhere

*TL;DR: [Whisper Caption](https://apps.apple.com/us/app/whisper-caption/id6758373934?ct=kevinmd-202607) transcribes and captions video entirely on the iPhone—OpenAI's Whisper model running locally, 90+ languages, karaoke-style word highlighting. Your footage never touches a server, and there's no subscription.*

Every video-captioning app I tried worked the same way: upload your video to their cloud, wait, get captions back, and pay monthly for the privilege. For a creator posting daily clips, maybe fine. But a lot of video shouldn't leave the phone—client work under NDA, family videos, an unreleased demo. And the subscription math is backwards: captioning is a burst activity (you caption five videos one weekend, then nothing for a month), which is exactly the usage pattern subscriptions are designed to exploit.

The thing is, the hard part of captioning—the speech recognition—no longer needs a server. Whisper, OpenAI's open-source speech model, runs comfortably on an iPhone's neural engine. So I built the app I wanted: pick a video, the transcription runs on the device, captions appear, you style them and export. The network is not involved. That's the entire architecture.

<img src="/images/whisper-caption-add-captions.png" alt="Whisper Caption editing a video with karaoke-style captions on an iPhone" width="430" />

*The basic loop: pick a video, get captions, adjust, export.*

## Word-level timing is the fun part

Whisper doesn't just give you text—it gives you timing per word. Whisper Caption uses that for karaoke-style highlighting: each word lights up as it's spoken. It reads as much more "produced" than static caption blocks, and it's the style short-form video has trained everyone to expect.

<img src="/images/whisper-caption-edit-every-word.png" alt="Whisper Caption transcript editor showing editable caption segments and timing" width="430" />

*Every segment stays editable, because automatic captions always need a human escape hatch.*

Getting this right in Chinese was its own adventure. For spaceless scripts like Chinese and Japanese, Whisper returns one coarse timing per sentence, so my first version highlighted the *entire sentence* at once—technically karaoke, spiritually a PowerPoint. The fix was exploding sentences into per-character units with CJK-aware spacing. I mention this partly because it was a satisfying bug, and partly because Chinese support isn't a checkbox here—the app's intro script literally demos Mandarin with pinyin captions. With 90+ languages in the model, Spanish, Chinese, and English were the three I tested hardest.

<img src="/images/whisper-caption-no-cloud.png" alt="Whisper Caption screenshot showing no cloud required and on-device captioning" width="430" />

*No cloud is not a slogan here; the captioning path runs locally.*

## What it doesn't do

No cloud features, obviously—that's the point, but it means no team libraries or cross-device sync. And on-device transcription is bounded by your phone's chip: a newer iPhone chews through video quickly, an older one takes noticeably longer. I think that trade is worth making explicit rather than hiding: you're trading a little patience for footage that never leaves your hand and a price that isn't a subscription.

<img src="/images/whisper-caption-privacy-settings.png" alt="Whisper Caption settings showing local Whisper model choices and local AI enhancement options" width="430" />

*The settings screen is boring in the right way: local model choices, no account machinery.*

There are multiple caption styles—light, dark, gradient, shadow—and export back to your camera roll. One of the first App Store reviews called it a "cracked app," which remains my favorite piece of user feedback across everything I've shipped.

[Whisper Caption is on the App Store](https://apps.apple.com/us/app/whisper-caption/id6758373934?ct=kevinmd-202607)—free to try, no account, no upload.