Add captions to video on iPad (on-device AI, no upload)

Adding captions to video on iPad should be the easy case. The screen is big enough to actually read a transcript, the chip is faster than most laptops, and the whole point of working on an iPad is doing this kind of job away from the desk. Instead, the category treats the iPad as an afterthought: the serious caption tools are subscription web apps that want your footage uploaded before they do anything, and the native options are mostly phone apps stretched to fill the display, with caption editing that fights you at every tap. This guide covers the path that uses the hardware you paid for — accurate, styled, burned-in captions generated on-device on the iPad itself, with no upload and no monthly fee.

One scoping note before anything else, because it decides whether this is your tool: the workflow below produces captioned vertical clips — 1080×1920, the 9:16 spec TikTok, Reels, and Shorts share. If you need to subtitle a 16:9 YouTube master or produce an .srt sidecar file, that’s a different category of tool and this post will say so plainly rather than pretend. If you need captioned short-form out of your recordings — which is what most people typing this search actually want — the iPad is quietly the best device for it, and here’s why.

The four jobs hiding inside “add captions”

“Add captions to video” reads like one task. It’s four, and the tools that frustrate people usually do one of them badly:

1. Transcribe accurately. Every downstream step inherits the transcript’s mistakes.
2. Time the words to the speech. Captions that lag or lead the audio read as broken even when the text is perfect.
3. Style the text for a phone screen. Short-form plays muted in a feed; the captions carry the first two seconds, so they have to be big, animated, and legible over moving footage.
4. Burn the text into the pixels. Sidecar subtitles vanish the moment a clip is downloaded, stitched, or reposted. Text that has to survive the feed has to be in the file.

The web caption tools do all four — on their servers, on their meter, with your unreleased footage sitting on their infrastructure while you wait for the upload. On an iPad with an M-series chip, shipping a gigabyte of video to a data center so it can do what the tablet in your hands could do locally is mostly habit. The on-device path does the same four jobs without the upload, the queue, or the subscription.

How to add captions to video on iPad with Clipolette

Clipolette is a native iPadOS app built around the caption-and-clip loop. It installs on iPadOS 18 or later, transcribes with Apple’s system speech frameworks, and burns word-by-word animated captions into vertical clips. There’s no account and no upload — the pipeline runs on-device and works offline. The full pass:

1. Import the video. One file at a time, MP4 or MOV, from Photos or the Files picker. That covers camera footage, a podcast recording AirDropped from your Mac, a Zoom export downloaded through Safari, or a file on an external SSD plugged into an iPad Pro — anything Files can see. There’s no URL ingest; the file has to be on (or mounted by) the iPad.
2. Frame it vertical. Every export is 1080×1920, and landscape footage gets reframed on the way: Auto-Crop, Follow Subject for active speakers, Blurred Fill, or a manual pan and zoom. Safe Zones preview exactly what the TikTok, Reels, and Shorts interfaces will cover, so you can confirm your captions clear the buttons before export instead of after posting.
3. Let it transcribe. The transcript appears synced to the video, fully editable before anything burns in.
4. Fix the transcript — the step the iPad is built for. Tap any word to correct it. For mistakes that repeat — a guest’s name misheard the same way across a 40-minute interview — Find & Replace corrects it across every caption in the project in one pass, with Replace All and Undo. On the iPad’s screen you can see the whole transcript context while you do it, which is exactly what makes this pass miserable on a phone.
5. Pick a caption style. Five word-by-word highlighted animations ship — Yellow Pop, Bold White, Cyan Burst, Beast, and Minimal — with color and size controls on top. Pick one and keep it across your clips; caption consistency is part of a channel’s visual identity.
6. Export. The iPad renders a 1080×1920 H.264 MP4 with the captions burned in and hands it to the share sheet — save to Photos, save to a Files folder, or send it straight into the posting app. One clip at a time. On iPadOS 26, the export keeps rendering if you switch apps.

For one already-trimmed clip, the whole pass is a couple of minutes. For a long recording you want several captioned shorts from, captioning is one stage inside the bigger loop — the app splits the recording into scenes, you keep the moments you want (or the AI ranks them for you on eligible hardware), and every kept clip gets the caption treatment.

Install Clipolette from the App Store — free, every editing tool included — and run one real clip through it. Transcript accuracy on your own names and jargon is the test that matters, and it costs two minutes to run.

Why the iPad is the best caption-review device Apple makes

The compute in this pipeline is never the bottleneck — the review is. Reading a transcript against the video, catching the three words the model missed, and checking that the caption block clears the TikTok UI is editorial work, and the iPad’s form factor fits it better than either sibling:

Better than the iPhone because of the screen. Reviewing a transcript on a 6-inch display means scrolling context you can’t see. On an 11- or 13-inch iPad, the transcript and the video preview coexist at readable sizes, and tapping the exact word you mean to fix stops being a precision exercise. The iPhone captioning guide covers the same workflow for when the phone is what you have — it works, it’s just slower per correction.

Different from the Mac in a useful way. The Mac version of this workflow wins on keyboard-driven bulk correction, and if your captioning happens at a desk it’s the comfortable choice. But caption review is fundamentally a touch task — tap the word, fix it, scrub two seconds, tap the next — and it doesn’t need a desk. The iPad does it on the couch after the recording session, on the train, in the venue’s green room right after the talk. With Stage Manager you can park your show notes or the guest’s name spelling in Safari next to the app while you work.

The Apple Pencil and a keyboard both help, neither is required. Precise taps on individual caption words are pleasant with the Pencil; a Magic Keyboard makes retyping a mangled phrase faster. The workflow is fully usable with fingers alone.

This is also why “iPad as a slightly-bigger iPhone” tools get captioning wrong: they inherit the phone’s cramped review ergonomics on a device whose entire advantage is space.

Getting the transcript right before it burns in

Burned-in captions have one hard property: mistakes are permanent. Once the text is in the pixels, a wrong name means a re-export. So spend your attention where it pays, in this order:

Proper nouns first. Names, brands, product names, acronyms — this is where every speech model guesses and where a wrong guess costs you credibility. If the term repeats, fix it once with Find & Replace rather than chasing instances down the transcript.

Then the phrase-level misses. Fast speech, crosstalk, and swallowed word endings produce the occasional mangled phrase. Play the clip with the transcript visible and fix what reads wrong.

Then stop. Captions transcribe speech; they don’t have to be prose. Copyediting spoken filler into written formality makes the captions feel dubbed. If the filler itself is the problem, cut it from the audio instead — Remove Fillers takes out the um/uh families and Tighten Pauses compresses the dead air, and the captions stay in sync with the tightened cut. (The filler-removal guide covers those tools in depth; they work the same way on iPad.)

Upstream of all of it: clean audio. A decent mic in a quiet room does more for caption accuracy than any amount of downstream correction.

Captions in a second language

If part of your audience watches in another language, Clipolette translates the caption track through Apple’s Translation framework — 19+ languages. Each language pack downloads once from Apple, and after that translation runs offline like everything else. That’s the one network touch in the pipeline beyond the App Store itself. The practical move for bilingual audiences is one recording, two exports — original captions for one feed, translated captions for the other — without transcribing twice or paying per minute for either pass. The caption translation guide covers the mechanics and the review pass a translated track deserves; they’re identical on iPad.

What this replaces, and what it doesn’t

Routing the job honestly across your options:

Use a timeline editor when the deliverable is the long video. Subtitling a 16:9 YouTube upload, producing an .srt file, captioning to an accessibility spec — those are general-editor jobs, and Clipolette deliberately doesn’t do them. The output here is always a 9:16 clip with the text burned in.

Use a web tool when you need what Clipolette scopes out — heavily animated branded caption template libraries, team workspaces, auto-posting. Those tools charge monthly for a reason; if the reason applies to you, pay it.

Use the on-device app when the deliverable is captioned clips, regularly. The economics compound: no per-minute meter, no subscription, no upload time, and footage that never leaves the iPad — which for client work and unreleased material is not a nicety but the requirement. The app is free with every editing tool included; a single one-time $24.99 purchase removes the export watermark, and it covers iPhone, iPad, and Mac on the same Apple Account. Nothing else is paid.

For the broader argument about why local-first is the right architecture for this category on iPad — privacy, latency, and the meter problem — the on-device video AI breakdown makes the full case.

Requirements, stated plainly

• Captioning, styling, Find & Replace, translation, and reframing: any iPad on iPadOS 18 or later. None of it is gated on Apple Intelligence.
• AI clip picking (Find Best Moments) is the one gated feature: it runs on Apple’s on-device models and needs iPadOS 26 on an M-series iPad or the A17 Pro iPad mini. On other iPads the picker doesn’t run — scene detection still splits the recording and you choose keepers with the toggles. The caption workflow is identical either way.
• No account, no login, no telemetry. The pipeline runs on-device and offline; translation packs download once per language.

FAQ

Which iPads can add captions, and which need Apple Intelligence? Any iPad running iPadOS 18 or later can transcribe, caption, translate, and export — captioning is not gated on Apple Intelligence. The one gated feature is AI clip picking (Find Best Moments), which needs iPadOS 26 on an M-series iPad or the A17 Pro iPad mini. On other iPads you pick scenes manually and the caption workflow is identical.

Can I export an SRT file or caption a landscape video on iPad? No. Clipolette burns the captions into the pixels and exports exactly one format: 1080×1920 vertical (9:16) H.264 MP4. It is a short-form clip tool, not a subtitling utility — for an .srt sidecar or a captioned 16:9 master, use a general editor, and use Clipolette for the vertical clips you cut out of that master.

Do captions on iPad work offline, or does the video get uploaded? The video never leaves the iPad. Transcription runs on Apple’s system speech frameworks, there is no account and no login, and the whole caption pipeline works offline. The one network touch beyond the App Store is caption translation: each language pack downloads once from Apple, then translation works offline too.

The bottom line

If you need to add captions to video on iPad and the video is short-form — clips for TikTok, Reels, Shorts, or a muted LinkedIn feed — the on-device path is the fast one and the iPad is the right screen for it: transcribe locally, fix the names once with Find & Replace, pick one of five word-by-word styles, and export a 9:16 file with the text baked in, without your footage touching a server or your card touching a subscription. If the job is subtitling long 16:9 masters, use a general editor — that’s its category. Get Clipolette on the App Store, caption one real clip on your iPad, and judge the transcript on the names that matter to you. That two-minute test answers this search better than any comparison table.