Burnin
Transcription runs on the phone
Burnin does not send your video anywhere. Apple's on-device speech transcriber turns speech into text right on the iPhone, which is why captioning keeps working on a flight, underground, or with the data switched off.
Features
Everything Burnin does
Grouped the way a video happens: import it, transcribe it, edit it, export it. Everything on this page works with no account.
On the phone, with no signal
Language packs, downloaded once
The first time you transcribe in a language, its model downloads. After that it stays on the phone, and every video in that language transcribes with no connection at all.
The speaker model, downloaded once
Speaker diarization uses an on-device FluidAudio model, around 100MB, that downloads once and then runs fully offline for every video afterward.
Automatic language detection
Burnin detects which language is being spoken on its own. A manual picker is there for when you would rather choose, or when a video mixes languages in a way detection cannot resolve on its own.
Transcription and captioning
Speech to readable caption chunks
A raw transcript is automatically segmented into caption-length lines, so what appears on screen reads at a glance instead of running on as one long sentence.
Gap detection that checks before it retries
Burnin looks for spans of the video that likely got dropped during transcription. It only re-transcribes a span when an independent speaker-diarization pass confirms real speech overlaps it, so it never invents captions over silence.
Speaker changes, labeled automatically
Diarization works out who is speaking when. Every caption where the speaker changes gets a "- " prefix, so a two-person conversation reads like one without any manual labeling.
Both models run detection and confirmation on-device
Segmentation and gap detection need no network signal at all, and the confirmation step that decides whether to retry a gap runs on the same on-device diarization pass used for speaker labeling.
The caption editor
A full editor sheet for every caption
Tap a caption, or the add button floating over the list, to open a large editor sheet at full height from the first frame, with its own delete control and a "Done" to close it.
A new caption lands where you are
Add a caption and it is inserted at the video's current position, in chronological order alongside its neighbors, not always tacked onto the end of the list.
Drag-trim on a visual timeline
Each caption's start and end time is a handle on a timeline strip. Drag either edge to retime it against the video underneath, with a margin above and below every block so there is always room to start a pan.
A draggable playhead
Drag the playhead on the timeline to scrub the video directly, instead of only watching it play.
Neighbor-clamping
Dragging one caption's edge past its neighbor is not allowed. Captions cannot be made to overlap by accident.
Undo and redo, deep
Every text edit and every trim is on the undo stack, twenty to thirty steps back, so an experiment you don't like is one gesture from gone.
Auto-save, debounced
Edits are saved automatically roughly every 600ms as you make them. Closing the app mid-edit, or a call coming in, does not cost you the work.
Live preview with captions burned in
The video preview shows exactly what the exported file will look like, captions and all, updated as you edit.
Title and description, before you share
Written from your own transcript
Right after export, before the share sheet opens, Burnin offers a title and a short description with a couple of hashtags, generated from the captions you just edited, not re-transcribed from scratch.
Three tones to pick from
Neutral, balanced, or clickbait. Switching tone regenerates the title and description to match, and you can skip the screen entirely and export with just the video.
Apple Intelligence, on the device
Generation runs through Apple's on-device Foundation Models when available, so the transcript never leaves the phone to be summarized.
A fallback with no capability gate
On a device without Apple Intelligence, or if generation fails, an on-device heuristic takes over: it picks a named person, place, or organization from the transcript for the title, and the single most information-dense sentence for the description, never just the first line.
Crop and export
Four crop options
Original, 9:16 for Reels, TikTok and Shorts, 1:1 for square feed posts, or 16:9 widescreen. Pick the shape the destination actually wants.
Three export resolutions
720p, 1080p or 4K, with captions burned directly into the exported file rather than added as a separate track.
Captions burned in, not a subtitle file
The export is a single video file with the captions part of the picture, so it plays correctly anywhere without needing a subtitle track to be attached separately.
Projects
Every video becomes a project
Import a video and it becomes a project in your library, with its captions and edits kept alongside it.
Rename and delete
Rename a project to find it later, delete one you no longer need, or select several at once to clean up in bulk.
Recovers a missing source file
If the original video file goes missing locally, Burnin can recover it from your Photos library instead of leaving the project broken.
Stored on your phone, not a server
Projects are stored locally, on-device. There is no cloud sync and no account for them to belong to.
Requirements
| Platform | iPhone and iPad (not iPad-optimized yet), iOS 18.0 or later |
|---|---|
| On-device transcription | iPhone 12 and later. Older devices, and the Simulator, are not supported |
| Video formats | MP4, MOV |
| Max video length | 10 minutes |
| Account | None. Not required, not offered |
| Price | No ads, no subscription |
One thing captions are not
Automatic transcription is not perfect on any device, on any app. Burnin's editor exists because every caption is meant to be checked, not accepted blindly, and gap detection is a heuristic that reduces missed speech rather than a guarantee that none is ever missed.